paper_id
stringlengths 10
10
⌀ | yes_no
bool 2
classes | paper_index
int64 0
519
⌀ | evidence
stringlengths 0
37.7k
| question
stringlengths 4
11.7k
| answer
stringlengths 1
26k
|
|---|---|---|---|---|---|
null | false
| null |
What is the best way to get faster at running long distances?
|
To become faster at running requires practicing a lot of long slow distance running mixed in with occasional interval training. It's also important to avoid injuries by stretching, investing in good running shoes, and sleeping and eating well.
|
|
null | false
| null |
Are fresh vegetable brought at a grocery store fresher than frozen vegetables?
|
It's quite common for a lot of people to assume that fresh vegetables are fresher than frozen vegetables. In fact, many wrongly assume that freezing destroys nutrients. Nothing could be farther from the truth. To survive transportation and storage, fresh vegetables are often harvested prematurely at sub-par nutrition levels. On the other hand, frozen vegetables are harvested close to peak maturity and nutrition levels.
Unless you have your own kitchen garden or have access to a farmers market located close to where the vegetables are actually grown, frozen is a better bet - it's fresher, more convenient and rather crucially, more easily accessible for you to improve the health quotient of your meals. It's better to eat frozen peas all through the year than be puritanical about enjoying it only during winter. Eating healthily is as much a behavior science problem as it is a nutrition science problem, and it's easier to keep frozen peas around and eat it regularly than to try and remember what is currently in season and go out of the way to procure them fresh all the time. Long term convenience always trumps short term will power improvements.
|
|
null | false
| null |
Why are cats better than dogs in some people's opinion?
|
There are couple of reasons that people would prefer cats over dogs. The major one is that cats are lower maintenance than dogs. Most cats are happy indoors and doesn't need walks daily. This becomes a huge advantages as many younger generation people prefers more indoor activities today as well. There are couple other advances: 1. Cats are less likely to cause a true mess of your house than dogs due to their smaller size and weaker strength. 2. Cats are less smelly than dogs. 3. Cats are quieter than dogs as they don't bark or howl.
|
|
null | false
| null |
What are the words of House Allyrion?
|
"No Foe May Pass"
|
|
null | false
| null |
The Killers are an American rock band formed in Las Vegas in 2001 by Brandon Flowers (lead vocals, keyboards, bass) and Dave Keuning (lead guitar, backing vocals). After going through a number of short-term bass players and drummers in their early days, both Mark Stoermer (bass, rhythm guitar, backing vocals) and Ronnie Vannucci Jr. (drums, percussion) joined the band in 2002. The band's name is derived from a logo on the bass drum of a fictitious band portrayed in the music video for the New Order song "Crystal".
|
Is the killers a band
|
The Killers are an American rock band formed in Las Vegas in 2001 by Brandon Flowers (lead vocals, keyboards, bass) and Dave Keuning (lead guitar, backing vocals). After going through a number of short-term bass players and drummers in their early days, both Mark Stoermer (bass, rhythm guitar, backing vocals) and Ronnie Vannucci Jr. (drums, percussion) joined the band in 2002. The band's name is derived from a logo on the bass drum of a fictitious band portrayed in the music video for the New Order song Crystal.
|
null | false
| null |
Identify which instrument is string or percussion: Washboard, Schrammel gitarre
|
Schrammel gitarre is string, Washboard is percussion.
|
|
null | false
| 28
|
In this section we further analyze the performance of PARENT-W under different conditions, and compare to the other best metrics from Table TABREF37 .
To study the correlation as we vary the number of divergent references, we also collected binary labels from workers for whether a reference is entailed by the corresponding table. We define a reference as entailed when it mentions only information which can be inferred from the table. Each table and reference pair was judged by 3 independent workers, and we used the majority vote as the label for that pair. Overall, only INLINEFORM0 of the references were labeled as entailed by the table. Fleiss' INLINEFORM1 was INLINEFORM2 , which indicates a fair agreement. We found the workers sometimes disagreed on what information can be reasonably entailed by the table.
Figure FIGREF40 shows the correlations as we vary the percent of entailed examples in the evaluation set of WikiBio. Each point is obtained by fixing the desired proportion of entailed examples, and sampling subsets from the full set which satisfy this proportion. PARENT and RG-F remain stable and show a high correlation across the entire range, whereas BLEU and BLEU-T vary a lot. In the hyperparams category, the latter two have the worst correlation when the evaluation set contains only entailed examples, which may seem surprising. However, on closer examination we found that this subset tends to omit a lot of information from the tables. Systems which produce more information than these references are penalized by BLEU, but not in the human evaluation. PARENT overcomes this issue by measuring recall against the table in addition to the reference.
We check how different components in the computation of PARENT contribute to its correlation to human judgments. Specifically, we remove the probability INLINEFORM0 of an n-gram INLINEFORM1 being entailed by the table from Eqs. EQREF19 and EQREF23 . The average correlation for PARENT-W drops to INLINEFORM5 in this case. We also try a variant of PARENT with INLINEFORM6 , which removes the contribution of Table Recall (Eq. EQREF22 ). The average correlation is INLINEFORM7 in this case. With these components, the correlation is INLINEFORM8 , showing that they are crucial to the performance of PARENT.
BIBREF28 point out that hill-climbing on an automatic metric is meaningless if that metric has a low instance-level correlation to human judgments. In Table TABREF46 we show the average accuracy of the metrics in making the same judgments as humans between pairs of generated texts. Both variants of PARENT are significantly better than the other metrics, however the best accuracy is only INLINEFORM0 for the binary task. This is a challenging task, since there are typically only subtle differences between the texts. Achieving higher instance-level accuracies will require more sophisticated language understanding models for evaluation.
Each table and reference pair was judged by 3 independent workers, and we used the majority vote as the label for that pair.
|
How many workers were involved in the human judgment of each table and reference pair?
|
3 independent workers.
|
null | false
| null |
Why do people visit the national parks?
|
The national parks are a system of diverse natural and historical locations that are preserved for people to enjoy. Because there is such variety in the national parks there's sure to be something that interests you. Some parks are made up of vast natural landscapes and wildernesses that can be explored. Others are sanctuaries for wildlife where you can see animals in their natural habitat. Still other locations are on islands or have amazing coastline and includes underwater habitats and wildlife. While some national parks are harder to access, most can be enjoyed by the majority of people. If you hate paying high entrance fees to destination locations the national parks are also a great option. You can get an annual pass for less than $100 which allows you to visit all national parks and many national forests and monuments at no additional charge. However, if you want to stay within the national parks make sure to book well in advance as campsites and lodging fills up quickly. For the most popular parks you'll want to book six months in advance when the reservations open up. The national parks can be enjoyed by people of all ages and have stunning vistas you can drive right up to or week long backpacking trips for adventure seekers.
|
|
null | false
| null |
Why do little boys jump in a puddle of water or mud?
|
Because they can. Because it was there. Because they have a sense their mother or father will disapprove so they are naturally inclined to do so.
|
|
null | false
| null |
Write a preface to book where the author feels the need to properly caveat his story about warfare in the middle east, in case someone were to get offended by the content. Make sure the author conveys his deep respect for that region and all cultures represented therein.
|
Though the events described here may have some historic precedent, they are totally fictitious, as are equipment and its use. Similarly, any resemblance between the characters in this story and actual individuals would be totally coincidental. Finally, though the geographical locations are real, the characterization of certain countries is vaguely inspired by reality. One should not conclude that the author's views relative to any of these countries matches the views presented here. The entire Middle East has been a war cauldron for most of its history, and yet it remains both the birthplace of western civilization and the origin of wonderful scientific discoveries. Additionally, its contribution to the world of art, through the ago has been unquestioned. In short, whoever is painted as a villain today could turn out to be a hero tomorrow.
|
|
null | false
| 16
|
For unsupervised POS tagging, we use a Markov-structured syntax model in our approach, which is a popular structure for unsupervised tagging tasks BIBREF9 , BIBREF10 .
Following existing literature, we train and test on the entire WSJ corpus (49208 sentences, 1M tokens). We use 45 tag clusters, the number of POS tags that appear in WSJ corpus. We train the discrete HMM and the Gaussian HMM BIBREF9 as baselines. For the Gaussian HMM, mean vectors of Gaussian emissions are initialized with the empirical mean of all word vectors with an additive noise. We assume diagonal covariance matrix for INLINEFORM0 and initialize it with the empirical variance of the word vectors. Following BIBREF9 , the covariance matrix is fixed during training. The multinomial probabilities are initialized as INLINEFORM1 , where INLINEFORM2 . For our approach, we initialize the syntax model and Gaussian parameters with the pre-trained Gaussian HMM. The weights of layers in the rectified network are initialized from a uniform distribution with mean zero and a standard deviation of INLINEFORM3 , where INLINEFORM4 is the input dimension. We evaluate the performance of POS tagging with both Many-to-One (M-1) accuracy BIBREF23 and V-Measure (VM) BIBREF24 . Given a model we found that the tagging performance is well-correlated with the training data likelihood, thus we use training data likelihood as a unsupervised criterion to select the trained model over 10 random restarts after training 50 epochs. We repeat this process 5 times and report the mean and standard deviation of performance.
We compare our approach with basic HMM, Gaussian HMM, and several state-of-the-art systems, including sophisticated HMM variants and clustering techniques with hand-engineered features. The results are presented in Table TABREF32 . Through the introduced latent embeddings and additional neural projection, our approach improves over the Gaussian HMM by 5.4 points in M-1 and 5.6 points in VM. Neural HMM (NHMM) BIBREF10 is a baseline that also learns word representation jointly. Both their basic model and extended Conv version does not outperform the Gaussian HMM. Their best model incorporates another LSTM to model long distance dependency and breaks the Markov assumption, yet our approach still achieves substantial improvement over it without considering more context information. Moreover, our method outperforms the best published result that benefits from hand-engineered features BIBREF27 by 2.0 points on VM.
We found that most tagging errors happen in noun subcategories. Therefore, we do the one-to-one mapping between gold POS tags and induced clusters and plot the normalized confusion matrix of noun subcategories in Figure FIGREF35 . The Gaussian HMM fails to identify “NN” and “NNS” correctly for most cases, and it often recognizes “NNPS” as “NNP”. In contrast, our approach corrects these errors well.
Confusion Matrix. We found that most tagging errors happen in noun subcategories. Therefore, we do the one-to-one mapping between gold POS tags and induced clusters and plot the normalized confusion matrix of noun subcategories in Figure 4. The Gaussian HMM fails to identify “NN” and “NNS” correctly for most cases, and it often recognizes “NNPS” as “NNP”. In contrast, our approach corrects these errors well.
|
What errors does Gaussian HMM have in noun subcategories tagging?
|
The Gaussian HMM fails to identify “NN” and “NNS” correctly for most cases, and it often recognizes “NNPS” as “NNP”.
|
null | false
| null |
The French Revolution (French: Révolution française [ʁevɔlysjɔ̃ fʁɑ̃sɛːz]) was a period of radical political and societal change in France that began with the Estates General of 1789 and ended with the formation of the French Consulate in November 1799.
|
When was the french revolution
|
The French revolution started in 1789 and ended in November 1799
|
null | false
| 99
|
In this work we formulate our classification problem as follows: given two classes of news articles, respectively $D$ (disinformation) and $M$ (mainstream), a set of news articles $A_i$ and associated class labels $C_i \in \lbrace D,M\rbrace $, and a set of tweets $\Pi _i=\lbrace T_i^1, T_i^2, ...\rbrace $ each of which contains an Uniform Resource Locator (URL) pointing explicitly to article $A_i$, predict the class $C_i$ of each article $A_i$. There is huge debate and controversy on a proper taxonomy of malicious and deceptive information BIBREF1BIBREF2BIBREF15BIBREF16BIBREF17BIBREF3BIBREF11. In this work we prefer the term disinformation to the more specific fake news to refer to a variety of misleading and harmful information. Therefore, we follow a source-based approach, a consolidated strategy also adopted by BIBREF6BIBREF16BIBREF2BIBREF1, in order to obtain relevant data for our analysis. We collected:
Disinformation articles, published by websites which are well-known for producing low-credibility content, false and misleading news reports as well as extreme propaganda and hoaxes and flagged as such by reputable journalists and fact-checkers;
Mainstream news, referring to traditional news outlets which deliver factual and credible information.
We believe that this is currently the most reliable classification approach, but it entails obvious limitations, as disinformation outlets may also publish true stories and likewise misinformation is sometimes reported on mainstream media. Also, given the choice of news sources, we cannot test whether our methodology is able to classify disinformation vs factual but not mainstream news which are published on niche, non-disinformation outlets.
We collected tweets associated to a dozen US mainstream news websites, i.e. most trusted sources described in BIBREF18, with the Streaming API, and we referred to Hoaxy API BIBREF16 for what concerns tweets containing links to 100+ US disinformation outlets. We filtered out articles associated to less than 50 tweets. The resulting dataset contains overall $\sim $1.7 million tweets for mainstream news, collected in a period of three weeks (February 25th, 2019-March 18th, 2019), which are associated to 6,978 news articles, and $\sim $1.6 million tweets for disinformation, collected in a period of three months (January 1st, 2019-March 18th, 2019) for sake of balance of the two classes, which hold 5,775 distinct articles. Diffusion censoring effects BIBREF14 were correctly taken into account in both collection procedures. We provide in Figure FIGREF4 the distribution of articles by source and political bias for both news domains.
As it is reported that conservatives and liberals exhibit different behaviors on online social platforms BIBREF19BIBREF20BIBREF21, we further assigned a political bias label to different US outlets (and therefore news articles) following the procedure described in BIBREF2. In order to assess the robustness of our method, we performed classification experiments by training only on left-biased (or right-biased) outlets of both disinformation and mainstream domains and testing on the entire set of sources, as well as excluding particular sources that outweigh the others in terms of samples to avoid over-fitting.
For what concerns the Italian scenario we first collected tweets with the Streaming API in a 3-week period (April 19th, 2019-May 5th, 2019), filtering those containing URLs pointing to Italian official newspapers websites as described in BIBREF22; these correspond to the list provided by the association for the verification of newspaper circulation in Italy (Accertamenti Diffusione Stampa). We instead referred to the dataset provided by BIBREF23 to obtain a set of tweets, collected continuously since January 2019 using the same Twitter endpoint, which contain URLs to 60+ Italian disinformation websites. In order to get balanced classes (April 5th, 2019-May 5th, 2019), we retained data collected in a longer period w.r.t to mainstream news. In both cases we filtered out articles with less than 50 tweets; overall this dataset contains $\sim $160k mainstream tweets, corresponding to 227 news articles, and $\sim $100k disinformation tweets, corresponding to 237 news articles. We provide in Figure FIGREF5 the distribution of articles according to distinct sources for both news domains. As in the US dataset, we took into account censoring effects BIBREF14 by excluding tweets published before (left-censoring) or after two weeks (right-censoring) from the beginning of the collection process.
The different volumes of news shared on Twitter in the two countries are due both to the different population size of US and Italy (320 vs 60 millions) but also to the different usage of Twitter platform (and social media in general) for news consumption BIBREF24. Both datasets analyzed in this work are available from the authors on request.
A crucial aspect in our approach is the capability to fully capturing sharing cascades on Twitter associated to news articles. It has been reported BIBREF25 that the Twitter streaming endpoint filters out tweets matching a given query if they exceed 1% of the global daily volume of shared tweets, which nowadays is approximately $5\cdot 10^8$; however, as we always collected less than $10^6$ tweets per day, we did not incur in this issue and we thus gathered 100% of tweets matching our query.
We built Twitter diffusion networks following an approach widely adopted in the literature BIBREF6BIBREF17BIBREF2. We remark that there is an unavoidable limitation in Twitter Streaming API, which does not allow to retrieve true re-tweeting cascades because re-tweets always point to the original source and not to intermediate re-tweeting users BIBREF8BIBREF14; thus we adopt the only viable approach based on Twitter's public availability of data. Besides, by disentangling different interactions with multiple layers we potentially reduce the impact of this limitation on the global network properties compared to the single-layer approach used in our baseline.
Using the notation described in BIBREF26. we employ a multi-layer representation for Twitter diffusion networks. Sociologists have indeed recognized decades ago that it is crucial to study social systems by constructing multiple social networks where different types of ties among same individuals are used BIBREF27. Therefore, for each news article we built a multi-layer diffusion network composed of four different layers, one for each type of social interaction on Twitter platform, namely retweet (RT), reply (R), quote (Q) and mention (M), as shown in Figure FIGREF11. These networks are not necessarily node-aligned, i.e. users might be missing in some layers. We do not insert "dummy" nodes to represent all users as it would have severe impact on the global network properties (e.g. number of weakly connected components). Alternatively one may look at each multi-layer diffusion network as an ensemble of individual graphs BIBREF26; since global network properties are computed separately for each layer, they are not affected by the presence of any inter-layer edges.
In our multi-layer representation, each layer is a directed graph where we add edges and nodes for each tweet of the layer type, e.g. for the RT layer: whenever user $a$ retweets account $b$ we first add nodes $a$ and $b$ if not already present in the RT layer, then we build an edge that goes from $b$ to $a$ if it does not exists or we increment the weight by 1. Similarly for the other layers: for the R layer edges go from user $a$ (who replies) to user $b$, for the Q layer edges go from user $b$ (who is quoted by) to user $a$ and for the M layer edges go from user $a$ (who mentions) to user $b$.
Note that, by construction, our layers do not include isolated nodes; they correspond to "pure tweets", i.e. tweets which have not originated any interactions with other users. However, they are present in our dataset, and their number is exploited for classification, as described below.
We used a set of global network indicators which allow us to encode each network layer by a tuple of features. Then we simply concatenated tuples as to represent each multi-layer network with a single feature vector. We used the following global network properties:
Number of Strongly Connected Components (SCC): a Strongly Connected Component of a directed graph is a maximal (sub)graph where for each pair of vertices $u,v$ there is a path in each direction ($u\rightarrow v$, $v\rightarrow u$).
Size of the Largest Strongly Connected Component (LSCC): the number of nodes in the largest strongly connected component of a given graph.
Number of Weakly Connected Components (WCC): a Weakly Connected Component of a directed graph is a maximal (sub)graph where for each pair of vertices $(u, v)$ there is a path $u \leftrightarrow v$ ignoring edge directions.
Size of the Largest Weakly Connected Component (LWCC): the number of nodes in the largest weakly connected component of a given graph.
Diameter of the Largest Weakly Connected Component (DWCC): the largest distance (length of the shortest path) between two nodes in the (undirected version of) largest weakly connected component of a graph.
Average Clustering Coefficient (CC): the average of the local clustering coefficients of all nodes in a graph; the local clustering coefficient of a node quantifies how close its neighbours are to being a complete graph (or a clique). It is computed according to BIBREF28.
Main K-core Number (KC): a K-core BIBREF13 of a graph is a maximal sub-graph that contains nodes of internal degree $k$ or more; the main K-core number is the highest value of $k$ (in directed graphs the total degree is considered).
Density (d): the density for directed graphs is $d=\frac{|E|}{|V||V-1|}$, where $|E|$ is the number of edges and $|N|$ is the number of vertices in the graph; the density equals 0 for a graph without edges and 1 for a complete graph.
Structural virality of the largest weakly connected component (SV): this measure is defined in BIBREF14 as the average distance between all pairs of nodes in a cascade tree or, equivalently, as the average depth of nodes, averaged over all nodes in turn acting as a root; for $|V| > 1$ vertices, $SV=\frac{1}{|V||V-1|}\sum _i\sum _j d_{ij}$ where $d_{ij}$ denotes the length of the shortest path between nodes $i$ and $j$. This is equivalent to compute the Wiener's index BIBREF29 of the graph and multiply it by a factor $\frac{1}{|V||V-1|}$. In our case we computed it for the undirected equivalent graph of the largest weakly connected component, setting it to 0 whenever $V=1$.
We used networkx Python package BIBREF30 to compute all features. Whenever a layer is empty. we simply set to 0 all its features. In addition to computing the above nine features for each layer, we added two indicators for encoding information about pure tweets, namely the number T of pure tweets (containing URLs to a given news article) and the number U of unique users authoring those tweets. Therefore, a single diffusion network is represented by a vector with $9\cdot 4+2=38$ entries.
Aforementioned network properties can be qualitatively explained in terms of social footprints as follows: SCC correlates with the size of the diffusion network, as the propagation of news occurs in a broadcast manner most of the time, i.e. re-tweets dominate on other interactions, while LSCC allows to distinguish cases where such mono-directionality is somehow broken. WCC equals (approximately) the number of distinct diffusion cascades pertaining to each news article, with exceptions corresponding to those cases where some cascades merge together via Twitter interactions such as mentions, quotes and replies, and accordingly LWCC and DWCC equals the size and the depth of the largest cascade. CC corresponds to the level of connectedness of neighboring users in a given diffusion network whereas KC identifies the set of most influential users in a network and describes the efficiency of information spreading BIBREF17. Finally, d describes the proportions of potential connections between users which are actually activated and SV indicates whether a news item has gained popularity with a single and large broadcast or in a more viral fashion through multiple generations.
For what concerns different Twitter actions, users primarily interact with each other using retweets and mentions BIBREF20.
The former are the main engagement activity and act as a form of endorsement, allowing users to rebroadcast content generated by other users BIBREF31. Besides, when node B retweets node A we have an implicit confirmation that information from A appeared in B's Twitter feed BIBREF12. Quotes are simply a special case of retweets with comments.
Mentions usually include personal conversations as they allow someone to address a specific user or to refer to an individual in the third person; in the first case they are located at the beginning of a tweet and they are known as replies, otherwise they are put in the body of a tweet BIBREF20. The network of mentions is usually seen as a stronger version of interactions between Twitter users, compared to the traditional graph of follower/following relationships BIBREF32.
Therefore, we follow a source-based approach, a consolidated strategy also adopted by (Shao et al. 2018a)(Shao et al. 2016)(Bovet and Makse 2019)(Grinberg et al. 2019), in order to obtain relevant data for our analysis.
|
What approach is used to obtain relevant data for analysis?
|
It is a source-based approach and a consolidated strategy to obtain relevant data for analysis.
|
null | false
| null |
A large language model (LLM) is a language model consisting of a neural network with many parameters (typically billions of weights or more), trained on large quantities of unlabelled text using self-supervised learning. LLMs emerged around 2018 and perform well at a wide variety of tasks. This has shifted the focus of natural language processing research away from the previous paradigm of training specialized supervised models for specific tasks.
Properties
Though the term large language model has no formal definition, it often refers to deep learning models having a parameter count on the order of billions or more. LLMs are general purpose models which excel at a wide range of tasks, as opposed to being trained for one specific task (such as sentiment analysis, named entity recognition, or mathematical reasoning). The skill with which they accomplish tasks, and the range of tasks at which they are capable, seems to be a function of the amount of resources (data, parameter-size, computing power) devoted to them, in a way that is not dependent on additional breakthroughs in design.
Though trained on simple tasks along the lines of predicting the next word in a sentence, neural language models with sufficient training and parameter counts are found to capture much of the syntax and semantics of human language. In addition, large language models demonstrate considerable general knowledge about the world, and are able to "memorize" a great quantity of facts during training.
Hallucinations
Main article: Hallucination (artificial intelligence)
In artificial intelligence in general, and in large language models in particular, a "hallucination" is a confident response that does not seem to be justified by the model's training data.
Emergent abilities
On a number of natural language benchmarks involving tasks such as question answering, models perform no better than random chance until they reach a certain scale (in this case, measured by training computation), at which point their performance sharply increases. These are examples of emergent abilities.
Unpredictable abilities that have been observed in large language models but that were not present in simpler models (and that were not explicitly designed into the model) are usually called "emergent abilities". Researchers note that such abilities "cannot be predicted simply by extrapolating the performance of smaller models". These abilities are discovered rather than programmed-in or designed, in some cases only after the LLM has been publicly deployed. Hundreds of emergent abilities have been described. Examples include multi-step arithmetic, taking college-level exams, identifying the intended meaning of a word, chain-of-thought prompting, decoding the International Phonetic Alphabet, unscrambling a word’s letters, identifying offensive content in paragraphs of Hinglish (a combination of Hindi and English), and generating a similar English equivalent of Kiswahili proverbs.
Architecture and training
Large language models have most commonly used the transformer architecture, which, since 2018, has become the standard deep learning technique for sequential data (previously, recurrent architectures such as the LSTM were most common). LLMs are trained in an unsupervised manner on unannotated text. A left-to-right transformer is trained to maximize the probability assigned to the next word in the training data, given the previous context. Alternatively, an LLM may use a bidirectional transformer (as in the example of BERT), which assigns a probability distribution over words given access to both preceding and following context. In addition to the task of predicting the next word or "filling in the blanks", LLMs may be trained on auxiliary tasks which test their understanding of the data distribution such as Next Sentence Prediction (NSP), in which pairs of sentences are presented and the model must predict whether they appear side-by-side in the training corpus.
The earliest LLMs were trained on corpora having on the order of billions of words. The first model in OpenAI's GPT series was trained in 2018 on BookCorpus, consisting of 985 million words. In the same year, BERT was trained on a combination of BookCorpus and English Wikipedia, totalling 3.3 billion words. In the years since then, training corpora for LLMs have increased by orders of magnitude, reaching up to hundreds of billions or trillions of tokens.
LLMs are computationally expensive to train. A 2020 study estimated the cost of training a 1.5 billion parameter model (1-2 orders of magnitude smaller than the state of the art at the time) at $1.6 million.
A 2020 analysis found that neural language models' capability (as measured by training loss) increased smoothly in a power law relationship with number of parameters, quantity of training data, and computation used for training. These relationships were tested over a wide range of values (up to seven orders of magnitude) and no attenuation of the relationship was observed at the highest end of the range (including for network sizes up to trillions of parameters).
Application to downstream tasks
Between 2018 and 2020, the standard method for harnessing an LLM for a specific natural language processing (NLP) task was to fine tune the model with additional task-specific training. It has subsequently been found that more powerful LLMs such as GPT-3 can solve tasks without additional training via "prompting" techniques, in which the problem to be solved is presented to the model as a text prompt, possibly with some textual examples of similar problems and their solutions.
Fine-tuning
Main article: Fine-tuning (machine learning)
Fine-tuning is the practice of modifying an existing pretrained language model by training it (in a supervised fashion) on a specific task (e.g. sentiment analysis, named entity recognition, or part-of-speech tagging). It is a form of transfer learning. It generally involves the introduction of a new set of weights connecting the final layer of the language model to the output of the downstream task. The original weights of the language model may be "frozen", such that only the new layer of weights connecting them to the output are learned during training. Alternatively, the original weights may receive small updates (possibly with earlier layers frozen).
Prompting
See also: Prompt engineering and Few-shot learning (natural language processing)
In the prompting paradigm, popularized by GPT-3, the problem to be solved is formulated via a text prompt, which the model must solve by providing a completion (via inference). In "few-shot prompting", the prompt includes a small number of examples of similar (problem, solution) pairs. For example, a sentiment analysis task of labelling the sentiment of a movie review could be prompted as follows:
Review: This movie stinks.
Sentiment: negative
Review: This movie is fantastic!
Sentiment:
If the model outputs "positive", then it has correctly solved the task. In zero-shot prompting, no solve examples are provided. An example of a zero-shot prompt for the same sentiment analysis task would be "The sentiment associated with the movie review 'This movie is fantastic!' is".
Few-shot performance of LLMs has been shown to achieve competitive results on NLP tasks, sometimes surpassing prior state-of-the-art fine-tuning approaches. Examples of such NLP tasks are translation, question answering, cloze tasks, unscrambling words, and using a novel word in a sentence. The creation and optimisation of such prompts is called prompt engineering.
Instruction tuning
Instruction tuning is a form of fine-tuning designed to facilitate more natural and accurate zero-shot prompting interactions. Given a text input, a pretrained language model will generate a completion which matches the distribution of text on which it was trained. A naive language model given the prompt "Write an essay about the main themes of Hamlet." might provide a completion such as "A late penalty of 10% per day will be applied to submissions received after March 17." In instruction tuning, the language model is trained on many examples of tasks formulated as natural language instructions, along with appropriate responses. Various techniques for instruction tuning have been applied in practice. OpenAI's InstructGPT protocol involves supervised fine-tuning on a dataset of human-generated (prompt, response) pairs, followed by reinforcement learning from human feedback (RLHF), in which a reward function was learned based on a dataset of human preferences. Another technique, "self-instruct", fine-tunes the language model on a training set of examples which are themselves generated by an LLM (bootstrapped from a small initial set of human-generated examples).
https://en.wikipedia.org/wiki/Large_language_model
|
Given these paragraphs about Large language models, what are "emergent abilities"?
|
Unpredictable abilities that have been observed in large language models but that were not present in simpler models (and that were not explicitly designed into the model) are usually called "emergent abilities". Researchers note that such abilities "cannot be predicted simply by extrapolating the performance of smaller models". These abilities are discovered rather than programmed-in or designed, in some cases only after the LLM has been publicly deployed.
|
null | false
| 37
|
Mapudungun (iso 639-3: arn) is an indigenous language of the Americas spoken natively in Chile and Argentina, with an estimated 100 to 200 thousand speakers in Chile and 27 to 60 thousand speakers in Argentina BIBREF0. It is an isolate language and is classified as threatened by Ethnologue, hence the critical importance of all documentary efforts. Although the morphology of nouns is relatively simple, Mapudungun verb morphology is highly agglutinative and complex. Some analyses provide as many as 36 verb suffix slots BIBREF1. A typical complex verb form occurring in our corpus of spoken Mapudungun consists of five or six morphemes.
Mapudungun has several interesting grammatical properties. It is a polysynthetic language in the sense of BIBREF2; see BIBREF3 for explicit argumentation. As with other polysynthetic languages, Mapudungun has Noun Incorporation; however, it is unique insofar as the Noun appears to the right of the Verb, instead of to the left, as in most polysynthetic languages BIBREF4. One further distinction of Mapudungun is that, whereas other polysynthetic languages are characterized by a lack of infinitives, Mapudungun has infinitival verb forms; that is, while subordinate clauses in Mapudungun closely resemble possessed nominals and may occur with an analytic marker resembling possessor agreement, there is no agreement inflection on the verb itself. One further remarkable property of Mapudungun is its inverse voice system of agreement, whereby the highest agreement is with the argument highest in an animacy hierarchy regardless of thematic role BIBREF5.
Mapudungun (iso 639-3: arn) is an indigenous language of the Americas spoken natively in Chile and Argentina, with an estimated 100 to 200 thousand speakers in Chile and 27 to 60 thousand speakers in Argentina (Z ´u ˜niga, 2006, 41–3) .
|
What is the Mapudungun?
|
Mapudungun is an indigenous language of the Americas spoken natively in Chile and Argentina, with an estimated 100 to 200 thousand speakers in Chile and 27 to 60 thousand speakers in Argentina.
|
null | false
| null |
Summer is traditionally associated with hot or warm weather. In Mediterranean climates, it is also associated with dry weather, while in other places (particularly in Eastern Asia because of the monsoon) it is associated with rainy weather. The wet season is the main period of vegetation growth within the savanna climate regime. Where the wet season is associated with a seasonal shift in the prevailing winds, it is known as a monsoon.
In the northern Atlantic Ocean, a distinct tropical cyclone season occurs from 1 June to 30 November. The statistical peak of the Atlantic hurricane season is 10 September. The Northeast Pacific Ocean has a broader period of activity, but in a similar timeframe to the Atlantic. The Northwest Pacific sees tropical cyclones year-round, with a minimum in February and March and a peak in early September. In the North Indian basin, storms are most common from April to December, with peaks in May and November. In the Southern Hemisphere, the tropical cyclone season runs from the start of November until the end of April with peaks in mid-February to early March.
Thunderstorm season in the United States and Canada runs in the spring through summer but sometimes can run as late as October or even November in the fall. These storms can produce hail, strong winds and tornadoes, usually during the afternoon and evening.
|
When are storms most common in the North Indian basin?
|
In the North Indian basin, storms are most common from April to December, with peaks in May and November.
|
null | false
| null |
What are some board or card games that you recommend for ten year olds?
|
Some games that are popular among that demographic are "Uno", "Exploding Kittens", "War", and "Connect 4".
|
|
null | false
| null |
Classify these animals as either mammals, reptiles, birds, or other: king cobra, honey badger, whale, black kite, salmon, and French poodle
|
mammals: honey badger, whale, french poodle
reptiles: king cobra
birds: black kite
other: salmon
|
|
null | false
| null |
Kendall Boiler and Tank Company Building is a one-story commercial edifice located on 275 Third Street in Kendall Square, Cambridge, Massachusetts. The brick building was formerly owned by the Kendall Boiler and Tank Company and is part of the Blake and Knowles Steam Pump Company National Register District, on the National Register of Historic Places.
|
Give a list of proper nouns in the passage.
|
Kendall Boiler and Tank Company Building
Third Street
Kendall Square
Cambridge
Massachusetts
Kendall Boiler and Tank Company
Blake and Knowles Steam Pump Company National Register District
National Register of Historic Places
|
null | false
| null |
Long before any knowledge of electricity existed, people were aware of shocks from electric fish. Ancient Egyptian texts dating from 2750 BCE referred to these fish as the "Thunderer of the Nile", and described them as the "protectors" of all other fish. Electric fish were again reported millennia later by ancient Greek, Roman and Arabic naturalists and physicians. Several ancient writers, such as Pliny the Elder and Scribonius Largus, attested to the numbing effect of electric shocks delivered by electric catfish and electric rays, and knew that such shocks could travel along conducting objects. Patients with ailments such as gout or headache were directed to touch electric fish in the hope that the powerful jolt might cure them.
Ancient cultures around the Mediterranean knew that certain objects, such as rods of amber, could be rubbed with cat's fur to attract light objects like feathers. Thales of Miletus made a series of observations on static electricity around 600 BCE, from which he believed that friction rendered amber magnetic, in contrast to minerals such as magnetite, which needed no rubbing.
Thales was incorrect in believing the attraction was due to a magnetic effect, but later science would prove a link between magnetism and electricity. According to a controversial theory, the Parthians may have had knowledge of electroplating, based on the 1936 discovery of the Baghdad Battery, which resembles a galvanic cell, though it is uncertain whether the artifact was electrical in nature.
Electricity would remain little more than an intellectual curiosity for millennia until 1600, when the English scientist William Gilbert wrote De Magnete, in which he made a careful study of electricity and magnetism, distinguishing the lodestone effect from static electricity produced by rubbing amber. He coined the New Latin word electricus ("of amber" or "like amber",, elektron, the Greek word for "amber") to refer to the property of attracting small objects after being rubbed. This association gave rise to the English words "electric" and "electricity", which made their first appearance in print in Thomas Browne's Pseudodoxia Epidemica of 1646.
Further work was conducted in the 17th and early 18th centuries by Otto von Guericke, Robert Boyle, Stephen Gray and C. F. du Fay. Later in the 18th century, Benjamin Franklin conducted extensive research in electricity, selling his possessions to fund his work. In June 1752 he is reputed to have attached a metal key to the bottom of a dampened kite string and flown the kite in a storm-threatened sky. A succession of sparks jumping from the key to the back of his hand showed that lightning was indeed electrical in nature. He also explained the apparently paradoxical behavior of the Leyden jar as a device for storing large amounts of electrical charge in terms of electricity consisting of both positive and negative charges
In 1775, Hugh Williamson reported a series of experiments to the Royal Society on the shocks delivered by the electric eel; that same year the surgeon and anatomist John Hunter described the structure of the fish's electric organs. In 1791, Luigi Galvani published his discovery of bioelectromagnetics, demonstrating that electricity was the medium by which neurons passed signals to the muscles. Alessandro Volta's battery, or voltaic pile, of 1800, made from alternating layers of zinc and copper, provided scientists with a more reliable source of electrical energy than the electrostatic machines previously used. The recognition of electromagnetism, the unity of electric and magnetic phenomena, is due to Hans Christian Ørsted and André-Marie Ampère in 1819–1820. Michael Faraday invented the electric motor in 1821, and Georg Ohm mathematically analysed the electrical circuit in 1827. Electricity and magnetism (and light) were definitively linked by James Clerk Maxwell, in particular in his "On Physical Lines of Force" in 1861 and 1862.
While the early 19th century had seen rapid progress in electrical science, the late 19th century would see the greatest progress in electrical engineering. Through such people as Alexander Graham Bell, Ottó Bláthy, Thomas Edison, Galileo Ferraris, Oliver Heaviside, Ányos Jedlik, William Thomson, 1st Baron Kelvin, Charles Algernon Parsons, Werner von Siemens, Joseph Swan, Reginald Fessenden, Nikola Tesla and George Westinghouse, electricity turned from a scientific curiosity into an essential tool for modern life.
In 1887, Heinrich Hertz discovered that electrodes illuminated with ultraviolet light create electric sparks more easily. In 1905, Albert Einstein published a paper that explained experimental data from the photoelectric effect as being the result of light energy being carried in discrete quantized packets, energising electrons. This discovery led to the quantum revolution. Einstein was awarded the Nobel Prize in Physics in 1921 for "his discovery of the law of the photoelectric effect". The photoelectric effect is also employed in photocells such as can be found in solar panels.
The first solid-state device was the "cat's-whisker detector" first used in the 1900s in radio receivers. A whisker-like wire is placed lightly in contact with a solid crystal (such as a germanium crystal) to detect a radio signal by the contact junction effect. In a solid-state component, the current is confined to solid elements and compounds engineered specifically to switch and amplify it. Current flow can be understood in two forms: as negatively charged electrons, and as positively charged electron deficiencies called holes. These charges and holes are understood in terms of quantum physics. The building material is most often a crystalline semiconductor.
Solid-state electronics came into its own with the emergence of transistor technology. The first working transistor, a germanium-based point-contact transistor, was invented by John Bardeen and Walter Houser Brattain at Bell Labs in 1947, followed by the bipolar junction transistor in 1948.
|
What is transistor technology?
|
Solid-state electronics came into its own with the emergence of transistor technology. The first working transistor, a germanium-based point-contact transistor, was invented by John Bardeen and Walter Houser Brattain at Bell Labs in 1947, followed by the bipolar junction transistor in 1948.
|
null | false
| null |
Which is heavier: 1kg of flour or 1kg of steel?
|
A kilogram (kg) is a unit of weight, so it doesn't matter what material is being weighed - a kilogram will always be equal to a kilogram. Therefore, both weigh the same.
|
|
null | false
| null |
Rosa Louise McCauley Parks (February 4, 1913 – October 24, 2005) was an American activist in the civil rights movement best known for her pivotal role in the Montgomery bus boycott. The United States Congress has honored her as "the first lady of civil rights" and "the mother of the freedom movement". Parks became a NAACP activist in 1943, participating in several high profile civil rights campaigns. On December 1, 1955, in Montgomery, Alabama, Parks rejected bus driver James F. Blake's order to vacate a row of four seats in the "colored" section in favor of a White passenger, once the "White" section was filled. Parks was not the first person to resist bus segregation, but the National Association for the Advancement of Colored People (NAACP) believed that she was the best candidate for seeing through a court challenge after her arrest for civil disobedience in violating Alabama segregation laws, and she helped inspire the Black community to boycott the Montgomery buses for over a year. The case became bogged down in the state courts, but the federal Montgomery bus lawsuit Browder v. Gayle resulted in a November 1956 decision that bus segregation is unconstitutional under the Equal Protection Clause of the 14th Amendment to the U.S. Constitution.
Parks's act of defiance and the Montgomery bus boycott became important symbols of the movement. She became an international icon of resistance to racial segregation, and organized and collaborated with civil rights leaders, including Edgar Nixon and Martin Luther King Jr. At the time, Parks was employed as a seamstress at a local department store and was secretary of the Montgomery chapter of the NAACP. She had recently attended the Highlander Folk School, a Tennessee center for training activists for workers' rights and racial equality. Although widely honored in later years, she also suffered for her act; she was fired from her job, and received death threats for years afterwards. Shortly after the boycott, she moved to Detroit, where she briefly found similar work. From 1965 to 1988, she served as secretary and receptionist to John Conyers, an African-American US Representative. She was also active in the Black Power movement and the support of political prisoners in the US.
After retirement, Parks wrote her autobiography and continued to insist that there was more work to be done in the struggle for justice. Parks received national recognition, including the NAACP's 1979 Spingarn Medal, the Presidential Medal of Freedom, the Congressional Gold Medal, and a posthumous statue in the United States Capitol's National Statuary Hall. Upon her death in 2005, she was the first woman to lie in honor in the Capitol Rotunda. California and Missouri commemorate Rosa Parks Day on her birthday, February 4, while Ohio, Oregon, and Texas commemorate the anniversary of her arrest, December 1.
|
Who was Rosa Parks?
|
Rosa Louise McCauley Parks, born on February 4, 1913 was an American activist in the civil rights movement best known for her pivotal role in the Montgomery bus boycott. The United States Congress has honored her as "the first lady of civil rights" and "the mother of the freedom movement". Parks became a NAACP activist in 1943, participating in several high profile civil rights campaigns. On December 1, 1955, in Montgomery, Alabama, Parks rejected bus driver James F. Blake's order to vacate a row of four seats in the "colored" section in favor of a White passenger, once the "White" section was filled. Parks was not the first person to resist bus segregation, but the National Association for the Advancement of Colored People (NAACP) believed that she was the best candidate for seeing through a court challenge after her arrest for civil disobedience in violating Alabama segregation laws, and she helped inspire the Black community to boycott the Montgomery buses for over a year. The case became bogged down in the state courts, but the federal Montgomery bus lawsuit Browder v. Gayle resulted in a November 1956 decision that bus segregation is unconstitutional under the Equal Protection Clause of the 14th Amendment to the U.S. Constitution.
Parks's act of defiance and the Montgomery bus boycott became important symbols of the movement. She became an international icon of resistance to racial segregation, and organized and collaborated with civil rights leaders, including Edgar Nixon and Martin Luther King Jr. At the time, Parks was employed as a seamstress at a local department store and was secretary of the Montgomery chapter of the NAACP. She had recently attended the Highlander Folk School, a Tennessee center for training activists for workers' rights and racial equality. Although widely honored in later years, she also suffered for her act; she was fired from her job, and received death threats for years afterwards. Shortly after the boycott, she moved to Detroit, where she briefly found similar work. From 1965 to 1988, she served as secretary and receptionist to John Conyers, an African-American US Representative. She was also active in the Black Power movement and the support of political prisoners in the US.
After retirement, Parks wrote her autobiography and continued to insist that there was more work to be done in the struggle for justice. Parks received national recognition, including the NAACP's 1979 Spingarn Medal, the Presidential Medal of Freedom, the Congressional Gold Medal, and a posthumous statue in the United States Capitol's National Statuary Hall. Upon her death in 2005, she was the first woman to lie in honor in the Capitol Rotunda. California and Missouri commemorate Rosa Parks Day on her birthday, February 4, while Ohio, Oregon, and Texas commemorate the anniversary of her arrest, December 1.
|
null | false
| null |
The 2018 TicketGuardian 500 was a Monster Energy NASCAR Cup Series race held on March 11, 2018, at ISM Raceway in Avondale, Arizona. Contested over 312 laps on the one-mile (1.6 km) oval, it was the fourth race of the 2018 Monster Energy NASCAR Cup Series season. It would also be the last Phoenix race to be using the 2011 reconfiguration as the start/finish line will be moved to the back-straightaway just past turn two for the November race.
|
What is the 2018 TicketGuardian 500
|
The 2018 TicketGuardian 500 was a 312 lap NASCAR Cup Series race in Avondale, Arizona on March 11, 2018.
|
null | false
| null |
In addition to being well known for articles about scenery, history, and the most distant corners of the world, the magazine has been recognized for its book-like quality and its standard of photography. It was during the tenure of Society President Alexander Graham Bell and editor Gilbert H. Grosvenor (GHG) that the significance of illustration was first emphasized, in spite of criticism from some of the Board of Managers who considered the many illustrations an indicator of an “unscientific” conception of geography. By 1910, photographs had become the magazine's trademark and Grosvenor was constantly on the search for "dynamical pictures" as Graham Bell called them, particularly those that provided a sense of motion in a still image. In 1915, GHG began building the group of staff photographers and providing them with advanced tools including the latest darkroom.The magazine began to feature some pages of color photography in the early 1930s, when this technology was still in its early development. During the mid-1930s, Luis Marden (1913–2003), a writer and photographer for National Geographic, convinced the magazine to allow its photographers to use the so-called "miniature" 35 mm Leica cameras loaded with Kodachrome film over bulkier cameras with heavy glass plates that required the use of tripods.In 1959, the magazine started publishing small photographs on its covers, later becoming larger photographs. National Geographic photography quickly shifted to digital photography for both its printed magazine and its website. In subsequent years, the cover, while keeping its yellow border, shed its oak leaf trim and bare table of contents, to allow for a full page photograph taken for one of the month's articles. Issues of National Geographic are often kept by subscribers for years and re-sold at thrift stores as collectibles. The standard for photography has remained high over the subsequent decades and the magazine is still illustrated with some of the highest-quality photojournalism in the world. In 2006, National Geographic began an international photography competition, with over eighteen countries participating.
|
Extract all of the names of people in the below text separated by commas and ordered in alphabetical order by first name
|
Alexander Graham Bell, Gilbert H. Grosvenor, Luis Marden
|
null | false
| null |
What is Pascal?
|
Pascal is an imperative and procedural programming language, designed by Niklaus Wirth as a small, efficient language intended to encourage good programming practices using structured programming and data structuring. It is named in honour of the French mathematician, philosopher and physicist Blaise Pascal.
|
|
null | false
| null |
Heat transfer is a discipline of thermal engineering that concerns the generation, use, conversion, and exchange of thermal energy (heat) between physical systems. Heat transfer is classified into various mechanisms, such as thermal conduction, thermal convection, thermal radiation, and transfer of energy by phase changes. Engineers also consider the transfer of mass of differing chemical species (mass transfer in the form of advection), either cold or hot, to achieve heat transfer. While these mechanisms have distinct characteristics, they often occur simultaneously in the same system.
|
From the paragraph provided, extract the mechanisms of heat transfer.
|
The mechanisms of heat transfer include thermal conduction, thermal convection, thermal radiation, and transfer of energy by phase changes, although engineers also consider mass transfer in the form of advection a form of heat transfer.
|
null | false
| null |
What is the best-selling book of all time?
|
According to Guinness World Records as of 1995, the Bible is the best-selling book of all time with an estimated 5 billion copies sold and distributed.
|
|
null | false
| 104
|
As established in previous work BIBREF7 , BIBREF8 , we use a standard unsupervised SRL setup, consisting of the following steps:
The task we model, unsupervised semantic role induction, is the step 4 of this pipeline.
We select all the nonauxiliary verbs in a sentence as predicates.
|
Does the author select all the nonauxiliary verbs in a sentence as predicates?
|
Yes.
|
null | false
| null |
How do you make the color green out of other colors?
|
You can make green by mixing blue and yellow together
|
|
null | false
| null |
The Golden Retriever is a Scottish breed of retriever dog of medium size. It is characterised by a gentle and affectionate nature and a striking golden coat. It is commonly kept as a pet and is among the most frequently registered breeds in several Western countries. It is a frequent competitor in dog shows and obedience trials; it is also used as a gundog, and may be trained for use as a guide dog.
|
From the passage provided, extract the type of breed and the characteristics of the dog ?
|
It is a Scottish breed and it is characterised by a gentle and affectionate nature and also a golden coat.
|
null | false
| null |
Name a good ski mountain in Vermont
|
Jay Peak is an excellent mountain for skiing in VT. It sits at one of the most Northern parts of Vermont and is the Northern most ski resort in the state. Because of that, Jay Peak averages more snowfall then any other mountain in Vermont. Additionally, because of it's remote location it is also ideal for a getaway that feels like a true getaway. There is also an indoor waterpark, several excellent dining options, rock climbing wall, arcade, and a movie theater to keep the entire family entertained. Because of its proximity to Canada, you can also take a day trip to Montreal which is an added bonus. The ski conditions are also the best in Vermont which makes Jay Peak an ideal mountain for a ski vacation in Vermont
|
|
null | false
| 30
|
Tables TABREF15 and TABREF16 show the translation accuracy in BLEU scores, the INLINEFORM0 -value of the significance test by bootstrap resampling BIBREF4 and training time in hours until convergence. The encoder-decoder-reconstructor BIBREF3 requires slightly longer time to train than the baseline NMT, but we emphasize that decoding time remains the same with the encoder-decoder-reconstructor and baseline-NMT. The results show that the encoder-decoder-reconstructor BIBREF3 significantly improves translation accuracy by 1.01 points on ASPEC and 1.37 points on NTCIR in English-Japanese translation ( INLINEFORM1 ). However, it does not significantly improve translation accuracy in Japanese-English translation. In addition, it is proved that the encoder-decoder-reconstructor without pre-training worsens rather than improves translation accuracy.
Table TABREF24 shows examples of outputs of English-Japanese translations. In Example 1, “UTF8min乱 流 粘性 と 数値 粘性 の 大小 関係 により ,” (on the basis of the relation between turbulent viscosity and numerical viscosity in size) is missing in the output of baseline-NMT, but “UTF8min乱 流 粘性 と 数値 的 粘性 の 関係 を 基 に” (on the basis of the relation between turbulent viscosity and numerical viscosity) is present in the output of encoder-decoder-reconstructor. In Example 2, “UTF8min新生児” (newborn infant) and “UTF8min30歳以上の” (of 30 ‐ year ‐ old or more) are repeated in the output of baseline-NMT, but they appear only once in the output of encoder-decoder-reconstructor.
In addition, Figures FIGREF25 and FIGREF25 show the attention layer on baseline-NMT and encoder-decoder-reconstructor in each example. In Figure FIGREF25 , although the attention layer of baseline NMT attends input word “turbulent”, the decoder does not output “UTF8min乱流” (turbulent) but “UTF8min検討” (examined) at the 13th word. Thus, under-translation may be resulted from the hidden layer or the embedding layer instead of the attention layer. In Figure FIGREF25 , it is found that the attention layer of baseline-NMT repeatedly attends input words “newborn infant” and “30 ‐ year ‐ old or more”. Consequently, the decoder repeatedly outputs “UTF8min新生児” (newborn infant) and “UTF8min30歳以上の” (of 30 ‐ year ‐ old or more). On the other hand, the attention layer of encoder-decoder-reconstructor almost correctly attends input words.
Table TABREF28 shows a comparison of the number of word occurrences for each corpus and model. The columns show (i) the number of words that appear more frequently than the counterparts in the reference, and (ii) the number of words that appear more than once but are not included in the reference. Note that these numbers do not include unknown words, so (iii) shows the number of unknown words. In all the cases, the number of occurrence of redundant words is reduced in encoder-decoder-reconstructor. Thus, we confirmed that encoder-decoder-reconstructor achieves reduction of repeating and missing words while maintaining the quality of translation.
Thus, under-translation may be resulted from the hidden layer or the embedding layer instead of the attention layer.
|
Through experiments, what might be the cause of under-translation?
|
The hidden layer or the embedding layer.
|
null | false
| null |
The catalogue consists of a diverse range of astronomical objects, from star clusters and nebulae to galaxies. For example, Messier 1 is a supernova remnant, known as the Crab Nebula, and the great spiral Andromeda Galaxy is M31. Further inclusions followed, the first addition came from Nicolas Camille Flammarion in 1921, who added Messier 104 after finding Messier's side note in his 1781 edition exemplar of the catalogue. M105 to M107 were added by Helen Sawyer Hogg in 1947, M108 and M109 by Owen Gingerich in 1960, and M110 by Kenneth Glyn Jones in 1967.
|
From the passage provided, extract the years when the galaxies were added to the catalogue. Include the names of astronomers that added them. Separate them with a comma.
|
1921, Nicolas Camille Flammarion
1947, Helen Sawyer Hogg
1960, Owen Gingerich
1967, Kenneth Glyn Jones
|
null | false
| null |
Swedish citizenship can be acquired by naturalisation, also known as citizenship by application. A foreigner may be granted Swedish citizenship upon meeting certain requirements. Person should be able to prove identity, 18 years of age or older, has a permanent residence permit or
has a right of residence or residence card (applies to EU citizens or close relatives of EU citizens), meets the requirements for habitual residence and has maintained good conduct while in Sweden.
|
To acquire Swedish citizenship by naturalisation what should be the minimum age of the applicant and what documents are required.
|
Minimum age should be 18 years and applicant should have permanent residence permit or right of residence.
|
null | false
| null |
In chess, the fianchetto (English: /ˌfiənˈkɛtoʊ/ or /ˌfiənˈtʃɛtoʊ/; Italian: [fjaŋˈketto] "little flank") is a pattern of development wherein a bishop is developed to the second rank of the adjacent b- or g-file, the knight pawn having been moved one or two squares forward.
The fianchetto is a staple of many "hypermodern" openings, whose philosophy is to delay direct occupation of the centre with the plan of undermining and destroying the opponent's occupied centre. It also regularly occurs in Indian defences. The fianchetto is less common in Open Games (1.e4 e5), but the king bishop is sometimes fianchettoed by Black in the Ruy Lopez or by White in an uncommon variation of the Vienna Game.
One of the major benefits of the fianchetto is that it often allows the fianchettoed bishop to become more active. A fianchettoed position, however, also presents some opportunities for the opponent: if the fianchettoed bishop can be exchanged, the squares the bishop was formerly protecting will become weak (see hole) and can form the basis of an attack (particularly if the fianchetto was performed on the kingside). Exchanging the fianchettoed bishop should not be done lightly, therefore, especially if the enemy bishop on same-coloured squares is still on the board.
|
What is the main benefit of the fianchetto in chess openings?
|
The fianchetto follows the "hypermodern" opening philosophy. It lets a bishop become more active from the flanks, attacking the center of the board without directly occupying it.
|
null | false
| null |
Can we move away from fossil fuels in the next decade ?
|
According to the most recent report from the Intergovernmental Panel on Climate Change (IPCC) and a consensus statement from over 97% of the world’s climate scientists, burning fossil fuels for electricity, heat, and transportation is the largest contributor to human driven climate change. Fossil fuels are responsible for over 75% of Greenhouse gas emissions and over 97% of carbon dioxide emissions. These reports have also stressed that there’s an immediate need to move away from these fuels in the near future.
The challenge to this paradigm shift from fossil fuels to renewable energy is a combination of social and political barriers, the cost for new infrastructure to manage new energy portfolios , fossil fuel phase-out may lead to an increment in electricity prices and challenges to energy storage.
In April 2022, McKinsey published their Energy Insights Global Energy Perspective report, where they highlighted a rapid increase in traditional energy ( Oil and Gas) over the next 5 years, however the report also shows a rise in renewable energy and Hydrogen following that 5 year mark. While each adoption of this will differ in each country, the consensus is we will have a mixed energy portfolio (renewables and traditional) for the next 3 decades. The data also suggests as we see technology improve for more sustainable management of assets, batteries and resources, we will see a decline in the extraction , transportation and consumption of fossil fuels.
|
|
null | false
| null |
I like live music. What are some examples of improvisational live music?
|
There are many different types of improvisational live music. They range from Jazz music to the Spoken Word. There are many bands and performers who are famous for their improvisation on stage. Some examples are Charlie Parker, John Coltrane, the Grateful Dead, and Phish.
|
|
null | false
| null |
Which was the first satellite launched by China?
|
Dongfanghong-1 was the first satellite launched by China
|
|
null | false
| null |
What are three different club teams Cristiano Ronaldo has played for?
|
Cristiano Ronaldo has played for Manchester United, Real Madrid, and Juventus.
|
|
null | false
| 478
|
The goal of data augmentation for contrastive learning is to create realistically rational instances that maintain semantics through different transformation approaches. Unlike instances in vision and language domains, the underlying semantics of time series data is not recognizable to human, making it hard, if not impossible, to include human knowledge to data augmentation for time series data.
For example, rotating an image will not change its content or the label. While permuting a time series instance may ruin its signal patterns and generates a meaningless time series instance. In addition, the tremendous heterogeneity of real-life time series datasets further makes selections based on trial-and-errors impractical. Although multiple data augmentation methods have been proposed for time series data, there is less discussion on what is a good augmentation that is meaningful for a given learning task and dataset without prefabricated human priors. From our perspective, ideal data augmentations for contrastive representation should keep high fidelity, high variety, and adaptive to different datasets. The illustration and examples are shown in Figure.
High Fidelity. Augmentations with high fidelity maintain the semantic identity that is invariant to transformations. Considering the inexplicability in practical time series data, it is challenging to visually check the fidelity of augmentations. Thus, we assume that the semantic identity of a time series instance is presented by its label in the downstream task, which might be either available or unavailable during the training period. Here, we start our analysis from the supervised case and will extend it to the unsupervised case later. Based on the information theory, we define the objective that keeps high fidelity as the large mutual information (MI) between augmentation v and the label y, i.e., MI(v; y).
We consider augmentation v as a probabilistic function of x and a random variable , that v = g(x; ).
From the definition of mutual information, we have MI(v; y) = H(y) − H(y|v), where H(y) is the (Shannon) entropy of y and H(y|v) is the entropy of y conditioned on augmentation v. Since H(y) is irrelevant to data augmentations, the objective is equivalent to minimizing the conditional entropy H(y|v). Considering the efficient optimization, we follow and to approximate it with cross-entropy between y and ŷ, where ŷ is the prediction with augmentation v as the input and calculated via
where z is the representation and h w (•) is a prediction projector parameterized by w. The prediction projector is optimized by the classification objective. Then, the objective of high fidelity for supervised or semi-supervised cases is to minimize
where C is the number of labels.
In the unsupervised settings where y is unavailable, one-hot encoding y s ∈ R |X| is utilized as the pseudo label to replace y in Eq. (). The motivation is that augmented instances are still distinguishable from other instances with the classifier. We theoretically show that augmentations that preserving pseudo labels have the following properties.
Property 1 (Preserving Fidelity). If augmentation v preserves the one-hot encoding pseudo label, the mutual information between v and the downstream task label y (although not visible to training) is equivalent to that between raw input x and y, i.e., MI(v; y) = MI(x; y).
Property 2 (Adding New Information). By preserving the one-hot encoding pseudo label, augmentation v contains new information comparing to the raw input x, i.e., H(v) ≥ H(x).
Detailed proofs are shown in the Appendix A. These properties show that in the unsupervised setting, preserving the one-hot encoding pseudo label guarantees that the generated augmentations will not decrease the fidelity, regardless of the downstream tasks and variances inherent in the augmentations. Concurrently, it may introduce new information for contrastive learning.
Since the number of labels is equal to the number of instances in dataset X in an unsupervised case, direct optimization of Eq. () is inefficient and unscalable. Thus, we further relax it by approximating y with the batch-wise one-hot encoding y B , which decreases the number of labels C from the dataset size to the batch size.
High Variety. Sufficient variances in augmentations improve the generalization capacity of contrastive learning models. In the information theory, the uncertainty inherent in the random variable's possible outcomes is described by its entropy. Considering that augmented instances are generated based on the raw input x, we maximize the entropy of v conditioned on x, H(v|x), to maintain a high variety of augmentations. From the definition of conditional entropy, we have
We dismiss the first part since the unconstrained entropy of v can be dominated by meaningless noise. Considering the continuity of both v and x, we adopt a mutual information neural estimator, InfoNCE to approximately compute the mutual information for its practical effectiveness. Other MI estimators, such as Jensen-Shannon (JSD) estimator, normalized temperature-scaled cross-entropy (NT-Xent), and leaveone-out can also conveniently be the plug-and-play component in our framework. Then, the objective to encourage high variety is to maximize the InfoNCE between v and x:
x ∈X exp(sim(z x , z v ))
where v is an augmented instance of input instance x . z x , z v , and z v are representations of instance x, v, and v respectively. sim(z 1 , z 2 ) = z T 1 z 2 is the inner product of vectors z 1 and z 2 .
Criteria. Combining the information aware definition of both high fidelity and variety, we propose the criteria for selecting good augmentations without prior knowledge,
where β is a hyper-parameter to achieve the trade-off between fidelity and variety. Note that in the unsupervised settings, y is replaced by one-hot encoding pseudo label... Similar to the information bottleneck, InfoMin assumes that augmented views are functions of the input, which heavily constrains the variance of data augmentations. Besides, high fidelity property is dismissed in the unsupervised setting. It works for image datasets due to the availability of human knowledge. However, it may fail to generate reasonable augmentations for time series data.
Since the number of labels is equal to the number of instances in dataset X in an unsupervised case, direct optimization of Eq. (2) is inefficient and unscalable. Thus, we further relax it by approximating y with the batch-wise one-hot encoding yB, which decreases the number of labels C from the dataset size to the batch size.
|
Why is it impossible to use the label for each data point?
|
As we explained later in our paper, *"Since the number of labels is equal to the number of instances in dataset X in an unsupervised case, direct optimization of Eq. (2) is inefficient and unscalable. (Page 4).''* For example, if we have 10k data points, meaning that the pseudo label is a one-hot encoding with 10k elements, such that the output dimension of the classifier will be 10k, which is inefficient with poor scalability.
|
null | false
| null |
Classify the following as either muscle car or minivan: Odyssey, Mustang, Sienna, Camero, Challenger
|
Muscle car: Mustang, Camero, Challenger
Minivan: Odyssey, Sienna
|
|
null | false
| null |
Designed to supplement Disneyland in Anaheim, California, which had opened in 1955, the complex was developed by Walt Disney in the 1960s. "The Florida Project", as it was known, was intended to present a distinct vision with its own diverse set of attractions. Walt Disney's original plans also called for the inclusion of an "Experimental Prototype Community of Tomorrow" (EPCOT), a planned community intended to serve as a testbed for new city-living innovations. Walt Disney died on December 15, 1966, during the initial planning of the complex. After his death, the company wrestled with the idea of whether to bring the Disney World project to fruition; however, Walt's older brother, Roy O. Disney, came out of retirement to make sure Walt's biggest dream was realized. Construction started in 1967, with the company instead building a resort similar to Disneyland, abandoning the experimental concepts for a planned community. The Magic Kingdom was the first theme park to open in the complex, in 1971, followed by Epcot (1982), Disney's Hollywood Studios (1989), and Disney's Animal Kingdom (1998). It was Roy who insisted the name of the entire complex be changed from Disney World to Walt Disney World, ensuring that people would remember that the project was Walt's dream.
|
Given this paragraph, which was built first Disneyland or Disney World?
|
According to this text, Disneyland was built first in 1955, with Disney World construction starting later in 1967 and first opening in 1971.
|
null | false
| null |
Why is Arsenal FC a great football club?
|
Arsenal is one of the greatest football clubs in the world for many reasons. First of all, it has one of the largest global fanbases from all continents on the globe. The fan base headcount is within the top 10 of all existing sports teams. It has been one of the most successful clubs in terms of winning silverware, with 13 league titles and 14 FA cups to its name. It has a rich history of having international players that go on to become some of the greatest players in football/soccer history. The club is also located in the heart of London, making it a popular team for people from all around the world to come and visit. It has one of the best stadiums in Europe, Emirates Stadium, that can seat 65,000+ attendants and has an incredible atmosphere on game day.
|
|
null | false
| null |
What kind of species is the Linopyrga primitractus
|
Linopyrga primitractus is a species of sea snail, a marine gastropod mollusk in the family Pyramidellidae, the pyrams and their allies.
|
|
null | false
| null |
The Amazon rainforest, also called Amazon jungle or Amazonia, is a moist broadleaf tropical rainforest in the Amazon biome that covers most of the Amazon basin of South America. This basin encompasses 7,000,000 km2 (2,700,000 sq mi), of which 5,500,000 km2 (2,100,000 sq mi) are covered by the rainforest. This region includes territory belonging to nine nations and 3,344 formally acknowledged indigenous territories.
The majority of the forest, 60%, is in Brazil, followed by Peru with 13%, Colombia with 10%, and with minor amounts in Bolivia, Ecuador, French Guiana, Guyana, Suriname, and Venezuela. Four nations have "Amazonas" as the name of one of their first-level administrative regions, and France uses the name "Guiana Amazonian Park" for French Guiana's protected rainforest area. The Amazon represents over half of the planet's remaining rainforests, and comprises the largest and most biodiverse tract of tropical rainforest in the world, with an estimated 390 billion individual trees in about 16,000 species.[
|
Where is Amazon rainforest located?
|
The Amazon jungle is located in South America. It is spanned across Brazil, Peru, Colombia, Bolivia and Ecuador
|
null | false
| null |
What wine is the Oregon Willamette Valley known for?
|
Oregon's Willamette Valley is known best for producing renowned Pinot Noirs.
|
|
null | false
| null |
Why did Christopher Columbus think he had sailed to Asia instead of America? What maps did he have at his disposal?
|
The reason behind Columbus' belief that he had reached Asia is depicted on a map of the Atlantic Ocean from 1474, created by Florentine mathematician, astronomer, and cosmographer Paolo dal Pozzo Toscanelli. This map was later included in a literary and historical atlas of America by Bartholomew in 1884, with light blue areas marking regions that were not originally present on the map.
Toscanelli served as an inspiration for Columbus, who received a letter from him on June 25, 1474. The letter was sent to canon Fernam Martins in Lisbon, who consulted Toscanelli on behalf of King Alfonso V about the possibility of reaching India through a western route.
The map clearly shows that Columbus was both inspired and misguided by Toscanelli's plan, which underestimated the Earth's circumference. This miscalculation led Columbus to believe that he had reached Japan instead of a completely new and unfamiliar continent lying between Europe and Asia.
Toscanelli's proposed plan involved sailing west along the equator to reach India and the wealthy countries of East Asia. The map that Toscanelli enclosed, which is now lost, included the western fringes of Europe and Africa, as well as the eastern fringes of Asia, with the Atlantic Ocean in between. However, Toscanelli mistakenly believed that the distance was only one-third of the Earth's circumference, far less than the actual distance.
In a letter, Toscanelli writes, among other things:
I have already spoken once before of a route across the sea to the land of spices, much shorter than that of the Portuguese across Guinea. That is why His Highness the King is now asking me for information on this subject, or rather for an explanation so clear that even people without any scientific background can understand that this path must exist. Although I know that it has its basis in the spherical shape of the earth, I have nevertheless, in order to make myself better understood and facilitate the project, decided to prove that such a route exists by means of a specially constructed sea chart. I am therefore sending his highness a map drawn by myself, on which I have marked the coasts and islands which can be used as points of departure when one undertakes such a journey and on this journey keeps a steady westerly course
|
|
null | false
| null |
Provide a number of different ways that someone can digest content from a book
|
There's more than one way to digest content from a book. In addition to actually reading the book, you can listen the audio book, read a summary or review of the book, or listen to a podcast discussing the book. Sometimes it's also possible to watch a movie based on the book.
|
|
null | false
| 126
|
Automatic summarization has enjoyed wide popularity in natural language processing due to its potential for various information access applications. Examples include tools which aid users navigate and digest web content (e.g., news, social media, product reviews), question answering, and personalized recommendation engines. Single document summarization — the task of producing a shorter version of a document while preserving its information content — is perhaps the most basic of summarization tasks that have been identified over the years (see BIBREF0 , BIBREF0 for a comprehensive overview).
Modern approaches to single document summarization are data-driven, taking advantage of the success of neural network architectures and their ability to learn continuous features without recourse to preprocessing tools or linguistic annotations. Abstractive summarization involves various text rewriting operations (e.g., substitution, deletion, reordering) and has been recently framed as a sequence-to-sequence problem BIBREF1 . Central in most approaches BIBREF2 , BIBREF3 , BIBREF4 , BIBREF5 , BIBREF6 , BIBREF7 is an encoder-decoder architecture modeled by recurrent neural networks. The encoder reads the source sequence into a list of continuous-space representations from which the decoder generates the target sequence. An attention mechanism BIBREF8 is often used to locate the region of focus during decoding.
Extractive systems create a summary by identifying (and subsequently concatenating) the most important sentences in a document. A few recent approaches BIBREF9 , BIBREF10 , BIBREF11 , BIBREF12 conceptualize extractive summarization as a sequence labeling task in which each label specifies whether each document sentence should be included in the summary. Existing models rely on recurrent neural networks to derive a meaning representation of the document which is then used to label each sentence, taking the previously labeled sentences into account. These models are typically trained using cross-entropy loss in order to maximize the likelihood of the ground-truth labels and do not necessarily learn to rank sentences based on their importance due to the absence of a ranking-based objective. Another discrepancy comes from the mismatch between the learning objective and the evaluation criterion, namely ROUGE BIBREF13 , which takes the entire summary into account.
In this paper we argue that cross-entropy training is not optimal for extractive summarization. Models trained this way are prone to generating verbose summaries with unnecessarily long sentences and redundant information. We propose to overcome these difficulties by globally optimizing the ROUGE evaluation metric and learning to rank sentences for summary generation through a reinforcement learning objective. Similar to previous work BIBREF9 , BIBREF11 , BIBREF10 , our neural summarization model consists of a hierarchical document encoder and a hierarchical sentence extractor. During training, it combines the maximum-likelihood cross-entropy loss with rewards from policy gradient reinforcement learning to directly optimize the evaluation metric relevant for the summarization task. We show that this global optimization framework renders extractive models better at discriminating among sentences for the final summary; a sentence is ranked high for selection if it often occurs in high scoring summaries.
We report results on the CNN and DailyMail news highlights datasets BIBREF14 which have been recently used as testbeds for the evaluation of neural summarization systems. Experimental results show that when evaluated automatically (in terms of ROUGE), our model outperforms state-of-the-art extractive and abstractive systems. We also conduct two human evaluations in order to assess (a) which type of summary participants prefer (we compare extractive and abstractive systems) and (b) how much key information from the document is preserved in the summary (we ask participants to answer questions pertaining to the content in the document by reading system summaries). Both evaluations overwhelmingly show that human subjects find our summaries more informative and complete.
Our contributions in this work are three-fold: a novel application of reinforcement learning to sentence ranking for extractive summarization; corroborated by analysis and empirical results showing that cross-entropy training is not well-suited to the summarization task; and large scale user studies following two evaluation paradigms which demonstrate that state-of-the-art abstractive systems lag behind extractive ones when the latter are globally trained.
corroborated by analysis and empirical results showing that cross-entropy training is not well-suited to the summarization task;
|
In this paper, which training is not well suited to the summarization task?
|
Corroborated by analysis and empirical results showing that cross-entropy training is not well suited to the summarization task.
|
null | false
| null |
What are 6 different reasons to go on vacation?
|
1. re-connect with you partner
2. spend time with your children
3. learn a new culture
4. address burnout you are feeling in your regular life
5. learn a new language
6. eat delicious food
|
|
null | false
| 162
|
BIBREF13 proposed a MP framework under which many of the recently introduced GNNs can be reformulated. MP consists in an aggregation phase followed by a combination phase BIBREF14. More precisely, let $G(V,E)$ be a graph, and let us consider $v \in V$. At time $t+1$, a message vector $\mathbf {m}_v^{t+1}$ is computed from the representations of the neighbors $\mathcal {N}(v)$ of $v$:
The new representation $\mathbf {h}^{t+1}_v$ of $v$ is then computed by combining its current feature vector $\mathbf {h}^{t}_v$ with the message vector $\mathbf {m}_v^{t+1}$:
Messages are passed for $T$ time steps. Each step is implemented by a different layer of the MP network. Hence, iterations correspond to network depth. The final feature vector $\mathbf {h}_v^T$ of $v$ is based on messages propagated from all the nodes in the subtree of height $T$ rooted at $v$. It captures both the topology of the neighborhood of $v$ and the distribution of the vertex representations in it.
If a graph-level feature vector is needed, e.g., for classification or regression, a READOUT pooling function, that must be invariant to permutations, is applied:
Next, we present the MP network we developed for document understanding.
We represent a document as a statistical word co-occurrence network BIBREF18, BIBREF19 with a sliding window of size 2 overspanning sentences. Let us denote that graph $G(V,E)$. Each unique word in the preprocessed document is represented by a node in $G$, and an edge is added between two nodes if they are found together in at least one instantiation of the window. $G$ is directed and weighted: edge directions and weights respectively capture text flow and co-occurrence counts.
$G$ is a compact representation of its document. In $G$, immediate neighbors are consecutive words in the same sentence. That is, paths of length 2 correspond to bigrams. Paths of length more than 2 can correspond either to traditional $n$-grams or to relaxed $n$-grams, that is, words that never appear in the same sentence but co-occur with the same word(s). Such nodes are linked through common neighbors.
Master node. Inspired by BIBREF3, our $G$ also includes a special document node, linked to all other nodes via unit weight bi-directional edges. In what follows, let us denote by $n$ the number of nodes in $G$, including the master node.
We formulate our AGGREGATE function as:
where $\mathbf {H}^t \in \mathbb {R}^{n \times d}$ contains node features ($d$ is a hyperparameter), and $\mathbf {A} \in \mathbb {R}^{n \times n}$ is the adjacency matrix of $G$. Since $G$ is directed, $\mathbf {A}$ is asymmetric. Also, $\mathbf {A}$ has zero diagonal as we choose not to consider the feature of the node itself, only that of its incoming neighbors, when updating its representation. Since $G$ is weighted, the $i^{th}$ row of $A$ contains the weights of the edges incoming on node $v_i$. $\mathbf {D} \in \mathbb {R}^{n \times n}$ is the diagonal in-degree matrix of $G$. MLP denotes a multi-layer perceptron, and $\mathbf {M}^{t+1} \in \mathbb {R}^{n \times d}$ is the message matrix.
The use of a MLP was motivated by the observation that for graph classification, MP neural nets with 1-layer perceptrons are inferior to their MLP counterparts BIBREF14. Indeed, 1-layer perceptrons are not universal approximators of multiset functions. Note that like in BIBREF14, we use a different MLP at each layer.
Renormalization. The rows of $\mathbf {D}^{-1}\mathbf {A}$ sum to 1. This is equivalent to the renormalization trick of BIBREF9, but using only the in-degrees. That is, instead of computing a weighted sum of the incoming neighbors' feature vectors, we compute a weighted average of them. The coefficients are proportional to the strength of co-occurrence between words. One should note that by averaging, we lose the ability to distinguish between different neighborhood structures in some special cases, that is, we lose injectivity. Such cases include neighborhoods in which all nodes have the same representations, and neighborhoods of different sizes containing various representations in equal proportions BIBREF14. As suggested by the results of an ablation experiment, averaging is better than summing in our application (see subsection SECREF30). Note that instead of simply summing/averaging, we also tried using GAT-like attention BIBREF11 in early experiments, without obtaining better results.
As far as our COMBINE function, we use the Gated Recurrent Unit BIBREF20, BIBREF21:
Omitting biases for readability, we have:
where the $\mathbf {W}$ and $\mathbf {U}$ matrices are trainable weight matrices not shared across time steps, $\sigma (\mathbf {x}) = 1/(1+\exp (-\mathbf {x}))$ is the sigmoid function, and $\mathbf {R}$ and $\mathbf {Z}$ are the parameters of the reset and update gates. The reset gate controls the amount of information from the previous time step (in $\mathbf {H}^t$) that should propagate to the candidate representations, $\tilde{\mathbf {H}}^{t+1}$. The new representations $\mathbf {H}^{t+1}$ are finally obtained by linearly interpolating between the previous and the candidate ones, using the coefficients returned by the update gate.
Interpretation. Updating node representations through a GRU should in principle allow nodes to encode a combination of local and global signals (low and high values of $t$, resp.), by allowing them to remember about past iterations. In addition, we also explicitly consider node representations at all iterations when reading out (see Eq. DISPLAY_FORM18).
After passing messages and performing updates for $T$ iterations, we obtain a matrix $\mathbf {H}^T \in \mathbb {R}^{n \times d}$ containing the final vertex representations. Let $\hat{G}$ be graph $G$ without the special document node, and matrix $\mathbf {\hat{H}}^T \in \mathbb {R}^{(n-1) \times d}$ be the corresponding representation matrix (i.e., $\mathbf {H}^T$ without the row of the document node).
We use as our READOUT function the concatenation of self-attention applied to $\mathbf {\hat{H}}^T$ with the final document node representation. More precisely, we apply a global self-attention mechanism BIBREF22 to the rows of $\mathbf {\hat{H}}^T$. As shown in Eq. DISPLAY_FORM17, $\mathbf {\hat{H}}^T$ is first passed to a dense layer parameterized by matrix $\mathbf {W}_A^T \in \mathbb {R}^{d \times d}$. An alignment vector $\mathbf {a}$ is then derived by comparing, via dot products, the rows of the output of the dense layer $\mathbf {Y}^T \in \mathbb {R}^{(n-1) \times d}$ with a trainable vector $\mathbf {v}^T \in \mathbb {R}^d$ (initialized randomly) and normalizing with a softmax. The normalized alignment coefficients are finally used to compute the attentional vector $\mathbf {u}^T \in \mathbb {R}^d$ as a weighted sum of the final representations $\mathbf {\hat{H}}^T$.
Note that we tried with multiple context vectors, i.e., with a matrix $\mathbf {V}^T$ instead of a vector $\mathbf {v}^T$, like in BIBREF22, but results were not convincing, even when adding a regularization term to the loss to favor diversity among the rows of $\mathbf {V}^T$.
Master node skip connection. $\mathbf {h}_G^T \in \mathbb {R}^{2d}$ is obtained by concatenating $\mathbf {u}^T$ and the final master node representation. That is, the master node vector bypasses the attention mechanism. This is equivalent to a skip or shortcut connection BIBREF23. The reason behind this choice is that we expect the special document node to learn a high-level summary about the document, such as its size, vocabulary, etc. (more details are given in subsection SECREF30). Therefore, by making the master node bypass the attention layer, we directly inject global information about the document into its final representation.
Multi-readout. BIBREF14, inspired by Jumping Knowledge Networks BIBREF12, recommend to not only use the final representations when performing readout, but also that of the earlier steps. Indeed, as one iterates, node features capture more and more global information. However, retaining more local, intermediary information might be useful too. Thus, instead of applying the readout function only to $t=T$, we apply it to all time steps and concatenate the results, finally obtaining $\mathbf {h}_G \in \mathbb {R}^{T \times 2d}$ :
In effect, with this modification, we take into account features based on information aggregated from subtrees of different heights (from 1 to $T$), corresponding to local and global features.
Through the successive MP iterations, it could be argued that MPAD implicitly captures some soft notion of the hierarchical structure of documents (words $\rightarrow $ bigrams $\rightarrow $ compositions of bigrams, etc.). However, it might be beneficial to explicitly capture document hierarchy. Hierarchical architectures have brought significant improvements to many NLP tasks, such as language modeling and generation BIBREF24, BIBREF25, sentiment and topic classification BIBREF26, BIBREF27, and spoken language understanding BIBREF28, BIBREF29. Inspired by this line of research, we propose several hierarchical variants of MPAD, detailed in what follows. In all of them, we represent each sentence in the document as a word co-occurrence network, and obtain an embedding for it by applying MPAD as previously described.
MPAD-sentence-att. Here, the sentence embeddings are simply combined through self-attention.
MPAD-clique. In this variant, we build a complete graph where each node represents a sentence. We then feed that graph to MPAD, where the feature vectors of the nodes are initialized with the sentence embeddings previously obtained.
MPAD-path. This variant is similar to the clique one, except that instead of a complete graph, we build a path according to the natural flow of the text. That is, two nodes are linked by a directed edge if the two sentences they represent follow each other in the document.
Note that we tried with multiple context vectors, i.e., with matrix VT instead of a vector vT, like in (Lin et al. 2017), but results were not convincing, even when adding a reg larization term to the loss to favor diversity among the ro of VT.
|
Will enhancing diversity among the ro of VT improve the performance of matrix VT?
|
No, they tried but the results were convincing.
|
2004.01694
| false
| null |
Our paper investigates three aspects of human MT evaluation, with a special focus on assessing human–machine parity: the choice of raters, the use of linguistic context, and the creation of reference translations. We focus on the data shared by BIBREF3, and empirically test to what extent changes in the evaluation design affect the outcome of the human evaluation. We find that for all three aspects, human translations are judged more favourably, and significantly better than MT, when we make changes that we believe strengthen the evaluation design. Based on our empirical findings, we formulate a set of recommendations for human MT evaluation in general, and assessing human–machine parity in particular. All of our data are made publicly available for external validation and further analysis.
In this study, we address three aspects that we consider to be particularly relevant for human evaluation of MT, with a special focus on testing human–machine parity: the choice of raters, the use of linguistic context, and the construction of reference translations.
We empirically test and discuss the impact of these factors on human evaluation of MT in Sections SECREF3–SECREF5. Based on our findings, we then distil a set of recommendations for human evaluation of strong MT systems, with a focus on assessing human–machine parity (Section SECREF6).
We focus on the data shared by BIBREF3, and empirically test to what extent changes in the evaluation design affect the outcome of the human evaluation.
In this study, we address three aspects that we consider to be particularly relevant for human evaluation of MT, with a special focus on testing human–machine parity: the choice of raters, the use of linguistic context, and the construction of reference translations.
We empirically test and discuss the impact of these factors on human evaluation of MT in Sections SECREF3–SECREF5.
|
What empricial investigations do they reference?
|
The answers are shown as follows:
* empirically test to what extent changes in the evaluation design affect the outcome of the human evaluation
|
null | false
| null |
Bruce Nodwell, OC (May 12, 1914 – January 20, 2006) was a Canadian inventor who invented the Nodwell 110, a multi-purpose two-tracked vehicle capable of traversing a wide variety of adverse terrain, including sand, mud, muskeg, swamp, and snow.
In 1970, he was made an Officer of the Order of Canada, Canada's highest civilian honor, "for his contribution to the opening of the Canadian North through his inventions and development of various types of tracked vehicles". A mountain in Antarctica "Nodwell Peaks" and a lake in NWT bear his name.
|
Based on the reference text about Bruce Nodwell, what did he create?
|
Bruce Nodwell invented the Nodwell 110, a two-tracked vehicle capable of traversing a wide variety of terrain.
|
null | false
| null |
How should I decorate my living room wall? Can you provide some tips and ideas?
|
First, you need to consider what type of style you wish to convey in the entire living room. This will dictate how you should decorate the living room wall. If you have a "statement piece" such as a large plant somewhere in your living room, design the wall as a counterpart to it.
You can decorate a living room wall with far more than a painting, although having a painting or other artwork be the central focus is not necessarily bad. Consider a variety of wall art including wall sculptures, mirrors, clocks, floating shelves, plants, and bright, bold colors.
Draw a few sketches out on paper or a computer program before you start to get a general idea of what you want to convey on the wall. And most importantly, have fun!
|
|
null | false
| null |
Do you prefer pancakes or waffles?
|
There is no question in my mind as to the answer to this question. I undoubtedly prefer waffles. Waffles offer a superior textural experience with their crispy edges and corners and fluffy insides. Waffles are easy to make and consistent in their appearance which offers a low-stress cooking experience. The cube pockets of a waffle offer the perfect home to copious amounts of butter and syrup, which is the ultimate reason they are much preferable to the measly pancake.
|
|
null | false
| 33
|
Lesk86 observed that concepts that are related should share more words in their respective definitions than concepts that are less connected. He was able to perform word sense disambiguation by identifying the senses of words in a sentence with the largest number of overlaps between their definitions. An overlap is the longest sequence of one or more consecutive words that occur in both definitions. BanerjeeP03 extended this idea to WordNet, but observed that WordNet glosses are often very short, and did not contain enough information to distinguish between multiple concepts. Therefore, they created a super–gloss for each concept by adding the glosses of related concepts to the gloss of the concept itself (and then finding overlaps).
PatwardhanP06 adapted this measure to second–order co–occurrence vectors. In this approach, a vector is created for each word in a concept's definition that shows which words co–occur with it in a corpus. These word vectors are averaged to create a single co-occurrence vector for the concept. The similarity between the concepts is calculated by taking the cosine between the concepts second–order vectors. LiuMPMP12 modified and extended this measure to be used to quantify the relatedness between biomedical and clinical terms in the UMLS. The work in this paper can be seen as a further extension of PatwardhanP06 and LiuMPMP12.
He was able to perform word sense disambiguation by identifying the senses of words in a sentence with the largest number of overlaps between their definitions.
|
How did Lesk perform word sense disambiguation?
|
By identifying the senses of words in a sentence with the largest number of overlaps between their definitions.
|
1701.00185
| false
| null |
The clustering performance is evaluated by comparing the clustering results of texts with the tags/labels provided by the text corpus. Two metrics, the accuracy (ACC) and the normalized mutual information metric (NMI), are used to measure the clustering performance BIBREF38 , BIBREF48 . Given a text INLINEFORM0 , let INLINEFORM1 and INLINEFORM2 be the obtained cluster label and the label provided by the corpus, respectively. Accuracy is defined as: DISPLAYFORM0
Two metrics, the accuracy (ACC) and the normalized mutual information metric (NMI), are used to measure the clustering performance BIBREF38 , BIBREF48 .
|
What were the evaluation metrics used?
|
The answers are shown as follows:
* accuracy
* normalized mutual information
|
2002.11893
| false
| null |
Database Construction: we crawled travel information in Beijing from the Web, including Hotel, Attraction, and Restaurant domains (hereafter we name the three domains as HAR domains). Then, we used the metro information of entities in HAR domains to build the metro database. For the taxi domain, there is no need to store the information. Instead, we can call the API directly if necessary.
Goal Generation: a multi-domain goal generator was designed based on the database. The relation across domains is captured in two ways. One is to constrain two targets that locate near each other. The other is to use a taxi or metro to commute between two targets in HAR domains mentioned in the context. To make workers understand the task more easily, we crafted templates to generate natural language descriptions for each structured goal.
Dialogue Collection: before the formal data collection starts, we required the workers to make a small number of dialogues and gave them feedback about the dialogue quality. Then, well-trained workers were paired to converse according to the given goals. The workers were also asked to annotate both user states and system states.
Dialogue Annotation: we used some rules to automatically annotate dialogue acts according to user states, system states, and dialogue histories. To evaluate the quality of the annotation of dialogue acts and states, three experts were employed to manually annotate dialogue acts and states for 50 dialogues. The results show that our annotations are of high quality. Finally, each dialogue contains a structured goal, a task description, user states, system states, dialogue acts, and utterances.
Database Construction: we crawled travel information in Beijing from the Web, including Hotel, Attraction, and Restaurant domains (hereafter we name the three domains as HAR domains). Then, we used the metro information of entities in HAR domains to build the metro database.
Goal Generation: a multi-domain goal generator was designed based on the database. The relation across domains is captured in two ways. One is to constrain two targets that locate near each other. The other is to use a taxi or metro to commute between two targets in HAR domains mentioned in the context.
Dialogue Collection: before the formal data collection starts, we required the workers to make a small number of dialogues and gave them feedback about the dialogue quality. Then, well-trained workers were paired to converse according to the given goals.
Dialogue Annotation: we used some rules to automatically annotate dialogue acts according to user states, system states, and dialogue histories.
|
How was the dataset collected?
|
The answers are shown as follows:
* Database Construction: we crawled travel information in Beijing from the Web, including Hotel, Attraction, and Restaurant domains (hereafter we name the three domains as HAR domains). Then, we used the metro information of entities in HAR domains to build the metro database.
* Goal Generation: a multi-domain goal generator was designed based on the database. The relation across domains is captured in two ways. One is to constrain two targets that locate near each other. The other is to use a taxi or metro to commute between two targets in HAR domains mentioned in the context.
* Dialogue Collection: before the formal data collection starts, we required the workers to make a small number of dialogues and gave them feedback about the dialogue quality. Then, well-trained workers were paired to converse according to the given goals. The workers were also asked to annotate both user states and system states.
* Dialogue Annotation: we used some rules to automatically annotate dialogue acts according to user states, system states, and dialogue histories.
|
null | false
| null |
A square has angles of 30 degrees, 90 degrees and 150 degrees. What is the last angle in the square?
|
All the angles in a square must be 90 degrees. Squares cannot have 30 degree or 150 degree angles
|
|
1612.09113
| false
| null |
FLOAT SELECTED: Figure 1: Our Hierarchical Network. In this network, junior tasks are supervised in lower layers, with an unsupervised task (Language Modeling) at the most senior layer.
FLOAT SELECTED: Figure 1: Our Hierarchical Network. In this network, junior tasks are supervised in lower layers, with an unsupervised task (Language Modeling) at the most senior layer.
|
How many supervised tasks are used?
|
The answers are shown as follows:
* two
|
null | false
| null |
What is the largest lake in the united states?
|
Lake Superior
|
|
null | false
| null |
Identify which instrument is string or woodwind: Bassoon, Mandolin
|
Mandolin is string, Bassoon is woodwind.
|
|
null | false
| null |
The dry roasting process employs either the batch or continuous method. In the batch method, peanuts are heated in large quantities in a revolving oven at about 800 °F (430 °C). Next, the peanuts in each batch are uniformly held and roasted in the oven at 320 °F (160 °C) for about 40 to 60 minutes. This method is good to use when the peanuts differ in moisture content. In the continuous method, a hot air roaster is employed. The peanuts pass through the roaster whilst being rocked to permit even roasting. A photometer indicates the completion of dry roasting. This method is favored by large manufacturers since it can lower the rate of spoilage and requires less labor.
|
Given this paragraph about roasting peanuts to make peanut butter, what temperature are peanuts roasted using the batch method in celsius?
|
Peanuts are heated in a revolving oven at about 430 °C using the batch method.
|
null | false
| null |
Why do humans seek meaning in their lives?
|
Every human life is a blend of joy and suffering. For some, joy comes more often. But for many, suffering can come to dominate their lives, due to abuse, the loss of loved ones, difficult relationships, mental illnesses, or disability. By finding meaning in ones life, the hard times full of suffering can be tempered with hope: for a better life for your children, for a better world, or a better after-life. Without finding something to give your life meaning, you risk falling into nihilism, anger, and despair. And those feelings tend to set you on a self-reinforcing path, and drag others down with you. While people often start looking for meaning in their lives when they encounter suffering to explain that suffering, it is in embracing a purpose that you can find hope and reduce the suffering in the world.
|
|
null | false
| null |
Dataphor is an open-source truly-relational database management system (RDBMS) and its accompanying user interface technologies, which together are designed to provide highly declarative software application development. The Dataphor Server has its own storage engine or it can be a virtual, or federated, DBMS, meaning that it can utilize other database engines for storage.
Dataphor has been praised for its adherence to relational principles, more closely so than any SQL product.
|
Give me a summary of Dataphor based on this text
|
Dataphor is an open-source database management system that provides a declarative software application development that has been praised for its adherence to relational principles. The Dataphor Server has its own storage engine but it can also utilize other database engines for storage that are virtual or federated DBMS.
|
null | false
| null |
What are the best sports movies?
|
The Rocky movies are quite popular, especially the original Rocky. 42, Facing the Giants, The Natural, Moneyball, Friday Night Lights, Remember the Titans, Hoosiers, Rudy, Tin Cup, The Blind Side, etc are also worthy of recognition. Many movie fans would also consider movies like Caddyshack, even though that's more of a Comedy, yet evolves around a golf match.
|
|
null | false
| null |
How large are elephant eggs?
|
Elephants do not lay eggs. They give birth to live young.
|
|
null | false
| null |
Who starred as Rocky Balboa
|
Sylvester Stallone
|
|
null | false
| null |
Budapest has architecturally noteworthy buildings in a wide range of styles and from distinct time periods, from the ancient times as Roman City of Aquincum in Óbuda (District III), which dates to around 89 AD, to the most modern Palace of Arts, the contemporary arts museum and concert hall.
Most buildings in Budapest are relatively low: in the early 2010s there were around 100 buildings higher than 45 metres (148 ft). The number of high-rise buildings is kept low by building legislation, which is aimed at preserving the historic cityscape and to meet the requirements of the World Heritage Site. Strong rules apply to the planning, authorisation and construction of high-rise buildings and consequently much of the inner city does not have any. Some planners would like see an easing of the rules for the construction of skyscrapers, and the possibility of building skyscrapers outside the city's historic core has been raised.
In the chronological order of architectural styles Budapest is represented on the entire timeline, starting with the Roman City of Aquincum representing ancient architecture.
The next determinative style is the Gothic architecture in Budapest. The few remaining Gothic buildings can be found in the Castle District. Buildings of note are no. 18, 20 and 22 on Országház Street, which date back to the 14th century and No. 31 Úri Street, which has a Gothic façade that dates back to the 15th century. Other buildings with Gothic features are the Inner City Parish Church, built in the 12th century, and the Mary Magdalene Church, completed in the 15th century. The most characteristic Gothic-style buildings are actually Neo-Gothic, like the most well-known Budapest landmarks, the Hungarian Parliament Building and the Matthias Church, where much of the original material was used (originally built in Romanesque style in 1015).
The next chapter in the history of human architecture is Renaissance architecture. One of the earliest places to be influenced by the Renaissance style of architecture was Hungary, and Budapest in particular. The style appeared following the marriage of King Matthias Corvinus and Beatrice of Naples in 1476. Many Italian artists, craftsmen and masons came to Buda with the new queen. Today, many of the original renaissance buildings disappeared during the varied history of Buda, but Budapest is still rich in renaissance and neo-renaissance buildings, like the famous Hungarian State Opera House, St. Stephen's Basilica and the Hungarian Academy of Sciences.
During the Turkish occupation (1541–1686), Islamic culture flourished in Budapest; multiple mosques and baths were built in the city. These were great examples of Ottoman architecture, which was influenced by Muslims from around the world including Turkish, Iranian, Arabian and to a larger extent, Byzantine architecture as well as Islamic traditions. After the Holy League conquered Budapest, they replaced most of the mosques with churches and minarets were turned into bell towers and cathedral spires. At one point the distinct sloping central square in Budapest became a bustling Oriental bazaar, which was filled with "the chatter of camel caravans on their way to Yemen and India". Budapest is in fact one of the few places in the world with functioning original Turkish bathhouses dating back to the 16th century, like Rudas Baths or Király Baths. Budapest is home to the northernmost place where the tomb of influential Islamic Turkish Sufi Dervish, Gül Baba is found. Various cultures converged in Hungary seemed to coalesce well with each other, as if all these different cultures and architecture styles are digested into Hungary's own way of cultural blend. A precedent to show the city's self-conscious is the top section of the city's main square, named as Szechenyi. When Turks came to the city, they built mosques here which was aggressively replaced with Gothic church of St. Bertalan. The rationale of reusing the base of the former Islamic building mosque and reconstruction into Gothic Church but Islamic style architecture over it is typically Islamic are still visible. An official term for the rationale is spolia. The mosque was called the djami of Pasha Gazi Kassim, and djami means mosque in Arabic. After Turks and Muslims were expelled and massacred from Budapest, the site was reoccupied by Christians and reformed into a church, the Inner City Parish Church (Budapest). The minaret and Turkish entranceway were removed. The shape of the architecture is its only hint of exotic past—"two surviving prayer niches facing Mecca and an ecumenical symbol atop its cupola: a cross rising above the Turkish crescent moon".
The most famous Budapest bridge, the Chain Bridge, the icon of the city's 19th century development, built in 1849
After 1686, the Baroque architecture designated the dominant style of art in catholic countries from the 17th century to the 18th century. There are many Baroque-style buildings in Budapest and one of the finest examples of preserved Baroque-style architecture is the Church of St. Anna in Batthyhány square. An interesting part of Budapest is the less touristy Óbuda, the main square of which also has some beautiful preserved historic buildings with Baroque façades. The Castle District is another place to visit where the best-known landmark Buda Royal Palace and many other buildings were built in the Baroque style.
The Classical architecture and Neoclassical architecture are the next in the timeline. Budapest had not one but two architects that were masters of the Classicist style. Mihály Pollack (1773–1855) and József Hild (1789–1867), built many beautiful Classicist-style buildings in the city. Some of the best examples are the Hungarian National Museum, the Lutheran Church of Budavár (both designed by Pollack) and the seat of the Hungarian president, the Sándor Palace. The most iconic and widely known Classicist-style attraction in Budapest is the Széchenyi Chain Bridge. Budapest's two most beautiful Romantic architecture buildings are the Great Synagogue in Dohány Street and the Vigadó Concert Hall on the Danube Promenade, both designed by architect Frigyes Feszl (1821–1884). Another noteworthy structure is the Budapest Western Railway Station, which was designed by August de Serres and built by the Eiffel Company of Paris in 1877.
The Hungarian Parliament, completed in 1904
Art Nouveau came into fashion in Budapest by the exhibitions which were held in and around 1896 and organised in connection with the Hungarian Millennium celebrations. Art Nouveau in Hungary (Szecesszió in Hungarian) is a blend of several architectural styles, with a focus on Hungary's specialities. One of the leading Art Nouveau architects, Ödön Lechner (1845–1914), was inspired by Indian and Syrian architecture as well as traditional Hungarian decorative designs. One of his most beautiful buildings in Budapest is the Museum of Applied Arts. Another examples for Art Nouveau in Budapest is the Gresham Palace in front of the Chain Bridge, the Hotel Gellért, the Franz Liszt Academy of Music or Budapest Zoo and Botanical Garden.
It is one of the world's outstanding urban landscapes and illustrates the great periods in the history of the Hungarian capital.
UNESCO
The second half of the 20th century also saw, under the communist regime, the construction of blocks of flats (panelház), as in other Eastern European countries. In the 21st century, Budapest faces new challenges in its architecture. The pressure towards the high-rise buildings is unequivocal among today's world cities, but preserving Budapest's unique cityscape and its very diverse architecture, along with green areas, is force Budapest to balance between them. The Contemporary architecture has wide margin in the city. Public spaces attract heavy investment by business and government also, so that the city has gained entirely new (or renovated and redesigned) squares, parks and monuments, for example the city central Kossuth Lajos square, Deák Ferenc square and Liberty Square. Numerous landmarks are created in the last decade in Budapest, like the National Theatre, Palace of Arts, Rákóczi Bridge, Megyeri Bridge, Budapest Airport Sky Court among others, and millions of square meters of new office buildings and apartments. But there are still large opportunities in real estate development in the city.
|
From the text provided, extract the list of styles found on Budapest architecture. Separate them with a semi-colon.
|
Gothic; Renaissance; Ottoman; Baroque; Classical; Neoclassical; Contemporary
|
null | false
| null |
Give me a quick exercise I can do every day
|
Choose a body-weight exercise of your choice - for example push ups, squats or lunges. Set a timer for 2 minutes. Do as many repetitions as you can before the 2 minutes are up. Take break as often as you need. Do this every day and track the number of repetitions.
|
|
null | false
| 151
|
- !`Socorro, me ha picado una víbora!
- ?`Cobra?
- No, gratis.[5]
Google Translation:
- Help, I was bitten by a snake!
- Does it charge?
- Not free.
[4]https://github.com/bfarzin/haha_2019_final, Accessed on 19 June 2019 [5]https://www.fluentin3months.com/spanish-jokes/, Accessed on 19 June 2019
Humor does not translate well because it often relies on double-meaning or a subtle play on word choice, pronunciation, or context. These issues are further exacerbated in areas where space is a premium (as frequent on social media platforms), often leading to usage and development of shorthand, in-jokes, and self-reference. Thus, building a system to classify the humor of tweets is a difficult task. However, with transfer-learning and the Fast.ai library, we can build a high quality classifier in a foreign language. Our system outperforms a Naive Bayes Support Vector Machine (NBSVM) baseline, which is frequently considered a "strong baseline" for many Natural Language Processing (NLP) related tasks (see Wang et al BIBREF0 ).
Rather than hand-crafted language features, we have taken an "end to end" approach building from the raw text to a final model that achieves the tasks as presented. Our paper lays out the details of the system and our code can be found in a GitHub repository for use by other researchers to extend the state of the art in sentiment analysis.
Our contributions are three fold. First, we apply transfer-learning of a language model based on a larger corpus of tweets. Second, we use a label smoothed loss, which provides regularization and allows full training of the final model without gradual unfreezing. Third, we select the best model for each task based on cross-validation and 20 random-seed initialization in the final network training step.
|
What are the contributions of this paper?
|
The contributions are three fold. First, the authors apply transfer-learning of a language model based on a larger corpus of tweets. Second, the authors use a label smoothed loss, which provides regularization and allows full training of the final model without gradual unfreezing. Third, the authors select the best model for each task based on cross-validation and 20 random-seed initialization in the final network training step.
|
null | false
| null |
What are top 3 cities to move in the US with no state tax
|
Most popular cities where people have been recently moving to among the states that don't have any state tax in the US are:
1. Miami, FL
2. Seattle, WA
3. Austin, TX
However, if you can tell me more in terms of what your professional interests and outdoor activities you like, I can tailor the response further
|
|
null | false
| 359
|
Word embeddings are low-dimensional, dense vector representations that capture semantic properties of words. Recently, they have gained tremendous popularity in Natural Language Processing (NLP) and have been used in tasks as diverse as text similarity BIBREF0 , part-of-speech tagging BIBREF1 , sentiment analysis BIBREF2 , and machine translation BIBREF3 . Although word embeddings are widely used across NLP, their stability has not yet been fully evaluated and understood. In this paper, we explore the factors that play a role in the stability of word embeddings, including properties of the data, properties of the algorithm, and properties of the words. We find that word embeddings exhibit substantial instabilities, which can have implications for downstream tasks.
Using the overlap between nearest neighbors in an embedding space as a measure of stability (see sec:definingStability below for more information), we observe that many common embedding spaces have large amounts of instability. For example, Figure FIGREF1 shows the instability of the embeddings obtained by training word2vec on the Penn Treebank (PTB) BIBREF4 . As expected, lower frequency words have lower stability and higher frequency words have higher stability. What is surprising however about this graph is the medium-frequency words, which show huge variance in stability. This cannot be explained by frequency, so there must be other factors contributing to their instability.
In the following experiments, we explore which factors affect stability, as well as how this stability affects downstream tasks that word embeddings are commonly used for. To our knowledge, this is the first study comprehensively examining the factors behind instability.
To our knowledge, this is the first study comprehensively examining the factors behind instability.
|
Is it the first study comprehensively examining the factors behind instability?
|
Yes.
|
null | false
| null |
Paleontology lies on the border between biology and geology, but differs from archaeology in that it excludes the study of anatomically modern humans. It now uses techniques drawn from a wide range of sciences, including biochemistry, mathematics, and engineering. Use of all these techniques has enabled paleontologists to discover much of the evolutionary history of life, almost all the way back to when Earth became capable of supporting life, nearly 4 billion years ago. As knowledge has increased, paleontology has developed specialised sub-divisions, some of which focus on different types of fossil organisms while others study ecology and environmental history, such as ancient climates.
|
Based on the following paragraph on paleontology, what's the difference between paleontology and archaeology?
|
Paleontology differs from archaeology in that archaeology includes the study of anatomically modern humans whereas paleontology does not.
|
null | false
| null |
How you can loose fat
|
1. Take a warm lemon water
2. Better take good breakfast and lunch
3. Do not take proper Dinner .
4. Finish all eating before 5 PM
5. No Suger
|
|
null | false
| null |
Which of the following vehicles are cars and which are trucks: Toyota Tacoma, Toyota Prius, Ford F-150, Honda Civic, Nissan Frontier, Hyundai Accent, Chevrolet Corvette, Mazda Miata
|
The Toyota Tacoma, Ford F-150 and Nissan Frontier are all trucks. The Toyota Prius, Honda Civic, Hyundai Accent, Chevrolet Corvette and Mazda Miata are all cars.
|
|
null | false
| null |
How is pickleball different from tennis?
|
Pickleball is easy to learn. Compared to tennis, it requires less strength and skills to play. It also takes up less space to set up a pickleball court. It does not create forearm tightness like tennis. It is a good sport for all ages.
|
|
1705.09665
| false
| null |
Our full dataset consists of all subreddits on Reddit from January 2013 to December 2014, for which there are at least 500 words in the vocabulary used to estimate our measures, in at least 4 months of the subreddit's history. We compute our measures over the comments written by users in a community in time windows of months, for each sufficiently active month, and manually remove communities where the bulk of the contributions are in a foreign language. This results in 283 communities ( INLINEFORM0 ), for a total of 4,872 community-months ( INLINEFORM1 ).
Our full dataset consists of all subreddits on Reddit from January 2013 to December 2014, for which there are at least 500 words in the vocabulary used to estimate our measures, in at least 4 months of the subreddit's history. We compute our measures over the comments written by users in a community in time windows of months, for each sufficiently active month, and manually remove communities where the bulk of the contributions are in a foreign language.
|
How did the select the 300 Reddit communities for comparison?
|
They selected all the subreddits from January 2013 to December 2014 with at least 500 words in the vocabulary and at least 4 months of the subreddit's history. They also removed communities with the bulk of the contributions are in foreign language.
|
null | false
| 78
|
Social media platforms have made the spreading of fake news easier, faster as well as able to reach a wider audience. Social media offer another feature which is the anonymity for the authors, and this opens the door to many suspicious individuals or organizations to utilize these platforms. Recently, there has been an increased number of spreading fake news and rumors over the web and social media BIBREF0. Fake news in social media vary considering the intention to mislead. Some of these news are spread with the intention to be ironic or to deliver the news in an ironic way (satirical news). Others, such as propaganda, hoaxes, and clickbaits, are spread to mislead the audience or to manipulate their opinions. In the case of Twitter, suspicious news annotations should be done on a tweet rather than an account level, since some accounts mix fake with real news. However, these annotations are extremely costly and time consuming – i.e., due to high volume of available tweets Consequently, a first step in this direction, e.g., as a pre-filtering step, can be viewed as the task of detecting fake news at the account level.
The main obstacle for detecting suspicious Twitter accounts is due to the behavior of mixing some real news with the misleading ones. Consequently, we investigate ways to detect suspicious accounts by considering their tweets in groups (chunks). Our hypothesis is that suspicious accounts have a unique pattern in posting tweet sequences. Since their intention is to mislead, the way they transition from one set of tweets to the next has a hidden signature, biased by their intentions. Therefore, reading these tweets in chunks has the potential to improve the detection of the fake news accounts.
In this work, we investigate the problem of discriminating between factual and non-factual accounts in Twitter. To this end, we collect a large dataset of tweets using a list of propaganda, hoax and clickbait accounts and compare different versions of sequential chunk-based approaches using a variety of feature sets against several baselines. Several approaches have been proposed for news verification, whether in social media (rumors detection) BIBREF0, BIBREF1, BIBREF2, BIBREF3, BIBREF4, or in news claims BIBREF5, BIBREF6, BIBREF7, BIBREF8. The main orientation in the previous works is to verify the textual claims/tweets but not their sources. To the best of our knowledge, this is the first work aiming to detect factuality at the account level, and especially from a textual perspective. Our contributions are:
[leftmargin=4mm]
We propose an approach to detect non-factual Twitter accounts by treating post streams as a sequence of tweets' chunks. We test several semantic and dictionary-based features together with a neural sequential approach, and apply an ablation test to investigate their contribution.
We benchmark our approach against other approaches that discard the chronological order of the tweets or read the tweets individually. The results show that our approach produces superior results at detecting non-factual accounts.
To the best of our knowledge, this is the first work aiming to detect factuality at the account level, and especially from a textual perspective.
|
Is this the first work aiming to detect factuality at the account level?
|
To the best of their knowledge, this is the first work aiming to detect factuality at the account level, and especially from a textual perspective.
|
null | false
| null |
Why do we say tip of the iceberg?
|
Majority of Icebergs size/mass tends to be submerged under water. So, their size is misleading when seen from above surface.
So the expression tip of the iceberg refers to a big issue when we see only a small part of it.
|
|
null | false
| null |
Dragon Ball Z is a Japanese anime television series produced by Toei Animation. Part of the Dragon Ball media franchise, it is the sequel to the 1986 Dragon Ball anime series and adapts the latter 325 chapters of the original Dragon Ball manga series created by Akira Toriyama. The series aired in Japan on Fuji TV from April 1989 to January 1996 and was later dubbed for broadcast in at least 81 countries worldwide.
Dragon Ball Z continues the adventures of Goku in his adult life as he and his companions defend the Earth against villains including aliens (Vegeta, Frieza), androids (Cell), and magical creatures (Majin Buu). At the same time, the story parallels the life of his son, Gohan, as well as the development of his rivals, Piccolo and Vegeta.
|
What is Dragon Ball Z?
|
Dragon Ball Z is a manga and an anime about an alien man who comes to Earth named Goku. It picks up the show from its prequel Dragon Ball and is analogous to Superman. Goku is a Saiyan, a race of fighting aliens, and as an adult, he and his friends, the Z fighters, defend the Earth from other Saiyans such as Raditz and Vegeta, alien overlords such as Frieza, androids such as Cell, and magical creatures such as Majin Buu.
|
null | false
| null |
Classify the following shapes as either two dimensional or three dimensional: cube, circle, sphere, triangle, cone, rhombus, square, and pyramid.
|
A cube, sphere, cone, and pyramid are three dimensional. A circle, triangle, rhombus, and square are two dimensional.
|
|
null | false
| 142
|
Machine translation has made remarkable progress, and studies claiming it to reach a human parity are starting to appear BIBREF0. However, when evaluating translations of the whole documents rather than isolated sentences, human raters show a stronger preference for human over machine translation BIBREF1. These findings emphasize the need to shift towards context-aware machine translation both from modeling and evaluation perspective.
Most previous work on context-aware NMT assumed that either all the bilingual data is available at the document level BIBREF2, BIBREF3, BIBREF4, BIBREF5, BIBREF6, BIBREF7, BIBREF8, BIBREF9, BIBREF10 or at least its fraction BIBREF11. But in practical scenarios, document-level parallel data is often scarce, which is one of the challenges when building a context-aware system.
We introduce an approach to context-aware machine translation using only monolingual document-level data. In our setting, a separate monolingual sequence-to-sequence model (DocRepair) is used to correct sentence-level translations of adjacent sentences. The key idea is to use monolingual data to imitate typical inconsistencies between context-agnostic translations of isolated sentences. The DocRepair model is trained to map inconsistent groups of sentences into consistent ones. The consistent groups come from the original training data; the inconsistent groups are obtained by sampling round-trip translations for each isolated sentence.
To validate the performance of our model, we use three kinds of evaluation: the BLEU score, contrastive evaluation of translation of several discourse phenomena BIBREF11, and human evaluation. We show strong improvements for all metrics.
We analyze which discourse phenomena are hard to capture using monolingual data only. Using contrastive test sets for targeted evaluation of several contextual phenomena, we compare the performance of the models trained on round-trip translations and genuine document-level parallel data. Among the four phenomena in the test sets we use (deixis, lexical cohesion, VP ellipsis and ellipsis which affects NP inflection) we find VP ellipsis to be the hardest phenomenon to be captured using round-trip translations.
Our key contributions are as follows:
we introduce the first approach to context-aware machine translation using only monolingual document-level data;
our approach shows substantial improvements in translation quality as measured by BLEU, targeted contrastive evaluation of several discourse phenomena and human evaluation;
we show which discourse phenomena are hard to capture using monolingual data only.
we show which discourse phenomena are hard to capture using monolingual data only.
|
What does the author find out about discourse phenomena in this paper?
|
Which discourse phenomena are hard to capture using monolingual data only.
|
null | false
| null |
Steps to choose right colour for occasion
|
Blue is the most recommended interview outfit
Warm colours like yellow, orange, or red are good for Company events
Bold colour like red or blue are good for parties
For court and legal the colours are blue, grey, and brown.
|
|
null | false
| null |
What is biathlon?
|
Biathlon is a winter sport that mixes nordic skying and shooting. Depending on the format, athletes ski a loop 3 or 5 times, and shoot 5 targets between each. The targets are 50 meters away. For each missed target, the participant will ski one 150 meter long penalty loop. The winner is the athlete who completes the 3 or 5 loops first.
|
|
null | false
| null |
Peru, officially the Republic of Peru (Spanish: República del Perú (help·info)), is a country in western South America. It is bordered in the north by Ecuador and Colombia, in the east by Brazil, in the southeast by Bolivia, in the south by Chile, and in the south and west by the Pacific Ocean. Peru is a megadiverse country with habitats ranging from the arid plains of the Pacific coastal region in the west to the peaks of the Andes mountains extending from the north to the southeast of the country to the tropical Amazon basin rainforest in the east with the Amazon River. Peru has a population of over 34 million, and its capital and largest city is Lima. At 1,285,216 km2 (496,225 sq mi), Peru is the 19th largest country in the world, and the third largest in South America.
|
Given this paragraph that describes the country of Peru, what is its population size?
|
Peru has a population of over 34 million people.
|
null | false
| null |
What are some ways to avoid social engineering attacks?
|
1. Educate yourself and your employees: Train yourself and your employees to recognize social engineering attacks and how to respond to them.
2. Be suspicious of unsolicited messages: If you receive a message that you were not expecting, be suspicious and verify the sender's identity before responding.
3. Use strong passwords and two-factor authentication: Use strong passwords and enable two-factor authentication to protect your accounts from unauthorized access.
4. Keep your software up to date: Keep your software and operating system up to date with the latest security patches to prevent vulnerabilities that attackers can exploit.
5. Use anti-virus and anti-malware software: Use anti-virus and anti-malware software to protect your devices from malicious software that can be used in social engineering attacks.
6. Be cautious of public Wi-Fi: Avoid using public Wi-Fi networks for sensitive activities, such as online banking or accessing confidential information.
|
|
null | false
| null |
ATV Offroad Fury 3 is a racing video game developed by Climax Racing and published by Sony Computer Entertainment for PlayStation 2.
The game was succeeded by ATV Offroad Fury 4 in 2006.
|
What video game console can you play ATV Offroad Fury 3 on?
|
ATV Offroad Fury 3 was published by Sony Computer Entertainment for the PlayStation 2 video game console.
|
null | false
| null |
Modern K-pop "idol" culture began in the 1990s, as K-pop idol music grew into a subculture that amassed enormous fandoms of teenagers and young adults. After a slump in early idol music, from 2003, TVXQ and BoA started a new generation of K-pop idols that broke the music genre into the neighboring Japanese market and continue to popularize K-pop internationally today. With the advent of online social networking services and South Korean TV shows, the current spread of K-pop and South Korean entertainment, known as the Korean Wave, is seen not only in East Asia and Southeast Asia, but also in Pakistan, Bangladesh, India, Latin America, North Africa, Southern Africa and East Africa, the Middle East and throughout the Western world, gaining an international audience.
|
How popular is K-pop around the world?
|
K-pop is a popular music genre that has a fan following across the whole world. It spread to worldwide fandom with the advent and proliferation of online social networking services.
|
null | false
| null |
What are three cities on the west coast of Florida?
|
Three cities on the west coast of Florida are Naples, Tampa, and Sarasota.
|
|
null | false
| null |
Who is Linkin Park
|
Linkin Park is an American rock band from Agoura Hills, California. The band's current lineup comprises vocalist/rhythm guitarist/keyboardist Mike Shinoda, lead guitarist Brad Delson, bassist Dave Farrell, DJ/turntablist Joe Hahn and drummer Rob Bourdon, all of whom are founding members. Vocalists Mark Wakefield and Chester Bennington are former members of the band. Categorized as alternative rock, Linkin Park's earlier music spanned a fusion of heavy metal and hip hop, while their later music features more electronica and pop elements.
|
|
null | false
| null |
Give me an idea of some cold beverages to drink on a hot day
|
Some cold beverages to drink on a hot day would be iced tea, lemonade, a milkshake, or water. If you enjoy alcohol, you could drink a refrigerated beer or a wine that is served cold, like sangria or rosé.
|
|
null | false
| null |
Foreverly (stylized as foreverly) is a collaborative album by Green Day singer/guitarist Billie Joe Armstrong and jazz/pop singer-songwriter Norah Jones. It was released on November 25, 2013, through Reprise Records.
Foreverly is a collection of traditional songs, and is a reinterpretation of the 1958 album Songs Our Daddy Taught Us by The Everly Brothers. It is promoted by the lead single "Long Time Gone". A lyric video for the song was released on November 13, 2013, and was followed by the release of a music video for "Silver Haired Daddy of Mine" five days later.
Background
|
What is Foreverly?
|
Foreverly is a remake of the album Songs Our Daddy Taught Us by The Everly Brothers. It was released in November 2013 and is a collaboration between Billie Joe Armstrong, of the band Green Day, and Norah Jones.
|
null | false
| null |
Who is the best soccer player of all time?
|
Pele is the biggest player of all time, he won three World Cup and he's the only player that has this achievement. In his career, he scored more than 1.280 goals in his career and won more than 60 championships playing with Santos Futebol Clube (Brazil), Cosmos (USA) and the Brazilian team.
|
|
null | false
| null |
What is a tardigrade?
|
A tardigrade, also called a water bear, is a microscopic animal with 8 legs. They can survive in a diverse range of conditions including outer space!
|
|
null | false
| null |
The First Law of thermodynamics is a formulation of the law of conservation of energy, adapted for thermodynamic processes. A simple formulation is: "The total energy in a system remains constant, although it may be converted from one form to another." Another common phrasing is that "energy can neither be created nor destroyed" (in a "closed system"). While there are many subtleties and implications that may be more precisely captured in more complex formulations, this is the essential principle of the First Law.
It distinguishes in principle two forms of energy transfer: heat, and thermodynamic work, for a system of a constant amount of matter. The law also defines the internal energy of a system, an extensive property for taking account of the balance of energies in the system.
|
From the passage provided, extract the two forms of energy transfer and list them in bullets.
|
The two forms of energy transfer are:
- Heat
- Thermodynamic work
|
null | false
| null |
The Cy Young Award is given annually to the best pitchers in Major League Baseball (MLB), one each for the American League (AL) and National League (NL). The award was first introduced in 1956 by Baseball Commissioner Ford Frick in honor of Hall of Fame pitcher Cy Young, who died in 1955. The award was originally given to the single best pitcher in the major leagues, but in 1967, after the retirement of Frick, the award was given to one pitcher in each league.
|
Given the reference text about the Cy Young Award, please describe what the award is and why it is awarded.
|
The Cy Young Award is given to the best pitches in the American and National leagues of Major League Baseball, for their achievements that exceed those of their peers.
|
null | false
| 151
|
- !`Socorro, me ha picado una víbora!
- ?`Cobra?
- No, gratis.[5]
Google Translation:
- Help, I was bitten by a snake!
- Does it charge?
- Not free.
[4]https://github.com/bfarzin/haha_2019_final, Accessed on 19 June 2019 [5]https://www.fluentin3months.com/spanish-jokes/, Accessed on 19 June 2019
Humor does not translate well because it often relies on double-meaning or a subtle play on word choice, pronunciation, or context. These issues are further exacerbated in areas where space is a premium (as frequent on social media platforms), often leading to usage and development of shorthand, in-jokes, and self-reference. Thus, building a system to classify the humor of tweets is a difficult task. However, with transfer-learning and the Fast.ai library, we can build a high quality classifier in a foreign language. Our system outperforms a Naive Bayes Support Vector Machine (NBSVM) baseline, which is frequently considered a "strong baseline" for many Natural Language Processing (NLP) related tasks (see Wang et al BIBREF0 ).
Rather than hand-crafted language features, we have taken an "end to end" approach building from the raw text to a final model that achieves the tasks as presented. Our paper lays out the details of the system and our code can be found in a GitHub repository for use by other researchers to extend the state of the art in sentiment analysis.
Rather than hand-crafted language features, we have taken an ”end to end” approach building from the raw text to a final model that achieves the tasks as presented.
|
How to build from the raw text to a final model that achieves the tasks?
|
The authors have taken an "end to end" approach building from the raw text to a final model that achieves the tasks as presented.
|
null | false
| null |
In 2002, McEwan discovered that he had a brother who had been given up for adoption during the Second World War; the story became public in 2007. The brother, a bricklayer named David Sharp, was born six years earlier than McEwan, when their mother was married to a different man. Sharp has the same mother and father as McEwan but was born from an affair that occurred before they married. After her first husband was killed in combat, McEwan's mother married her lover, and Ian was born a few years later. The brothers are in regular contact and McEwan has written a foreword to Sharp's memoir.
|
How do McEwan and Sharp share the same parents if their mother was married to a different man when Sharp was born?
|
While married to a man who would die in combat, their mother had an affair with a man (their father) she would later marry. That affair produced Sharp, and they proceeded to have another child, McEwan, following their marriage.
|
null | false
| null |
Apache Spark requires a cluster manager and a distributed storage system. For cluster management, Spark supports standalone (native Spark cluster, where you can launch a cluster either manually or use the launch scripts provided by the install package. It is also possible to run these daemons on a single machine for testing), Hadoop YARN, Apache Mesos or Kubernetes. For distributed storage, Spark can interface with a wide variety, including Alluxio, Hadoop Distributed File System (HDFS), MapR File System (MapR-FS), Cassandra, OpenStack Swift, Amazon S3, Kudu, Lustre file system, or a custom solution can be implemented. Spark also supports a pseudo-distributed local mode, usually used only for development or testing purposes, where distributed storage is not required and the local file system can be used instead; in such a scenario, Spark is run on a single machine with one executor per CPU core.
|
What are the main components of Apache Spark based on the paragraph?
|
Based on the paragraph, a cluster manager and a distributed storage are two main components of Apache Spark. For cluster manager, Spark supports a standalone cluster, or Hadoop YARN, Apache Mesos, or Kubernetes. For storage, Spark supports local file systems for development and testing, as well as Hadoop File System, Cassandra, and cloud storages
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.