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 |
Gandhi had dropped out of the cheapest college he could afford in Bombay. Mavji Dave Joshiji, a Brahmin priest and family friend, advised Gandhi and his family that he should consider law studies in London. In July 1888, his wife Kasturba gave birth to their first surviving son, Harilal. His mother was not comfortable about Gandhi leaving his wife and family, and going so far from home. Gandhi's uncle Tulsidas also tried to dissuade his nephew. Gandhi wanted to go. To persuade his wife and mother, Gandhi made a vow in front of his mother that he would abstain from meat, alcohol and women. Gandhi's brother Laxmidas, who was already a lawyer, cheered Gandhi's London studies plan and offered to support him. Putlibai gave Gandhi her permission and blessing.
On 10 August 1888, Gandhi aged 18, left Porbandar for Mumbai, then known as Bombay. Upon arrival, he stayed with the local Modh Bania community whose elders warned him that England would tempt him to compromise his religion, and eat and drink in Western ways. Despite Gandhi informing them of his promise to his mother and her blessings, he was excommunicated from his caste. Gandhi ignored this, and on 4 September, he sailed from Bombay to London, with his brother seeing him off. Gandhi attended University College, London, where he took classes in English literature with Henry Morley in 1888–1889.
He also enrolled at Inner Temple with the intention of becoming a barrister. His childhood shyness and self-withdrawal had continued through his teens. He retained these traits when he arrived in London, but joined a public speaking practice group and overcame his shyness sufficiently to practise law
|
Given this paragraph about Mahatma Gandhi, can you tell me why did he go to London?
|
Gandhi travelled to London in the late 1880s to study law. He first took English literature classes and then enrolled in Inner Temple with the intention to become a barrister.
|
null | false
| null |
Which is a species of fish? Tenpounder or Ninepounder
|
Tenpounder
|
|
1908.10383
| false
| null |
Building Extractive CNN/Daily Mail
Building Extractive CNN/Daily Mail
|
What is the problem with existing metrics that they are trying to address?
|
Answer with content missing: (whole introduction) However, recent
studies observe the limits of ROUGE and find in
some cases, it fails to reach consensus with human.
judgment (Paulus et al., 2017; Schluter, 2017).
|
null | false
| null |
Let me know if you think these are musical instruments or tools: Piano, chisel, Marimba, hammer, drill, scissor, violin, drums, screwdriver, brass, clarinet.
|
Piano, marimba, violin, drums and brass are all musical instruments. Piano belongs to the percussive chordophone family. A violin is a bowed stringed musical instrument. Marimba and the drums are percussion instruments. A brass falls under the category of wind instruments, while a clarinet is considered a woodwind instrument.
From your list, the following items are tools: chisel, hammer, drill and screwdriver. Chisel is a hand tool used for carving a hard material such as wood, stone, or metal. With a scissor, you normally cut paper and potentially some types of thin plastic. A hammer is a multi-purpose tool that you can use, for example, to secure objects with nails in wood, pull nails from wood and demolish objects. A drill is commonly used to drill holes or to drive screws into wood, metal, concrete and other materials. Last, a screwdriver is a mechanical hand tool that can you can use to drive screws in similar materials.
|
|
null | false
| 124
|
The task of speculation detection and scope resolution is critical in distinguishing factual information from speculative information. This has multiple use-cases, like systems that determine the veracity of information, and those that involve requirement analysis. This task is particularly important to the biomedical domain, where patient reports and medical articles often use this feature of natural language. This task is commonly broken down into two subtasks: the first subtask, speculation cue detection, is to identify the uncertainty cue in a sentence, while the second subtask: scope resolution, is to identify the scope of that cue. For instance, consider the example:
It might rain tomorrow.
The speculation cue in the sentence above is ‘might’ and the scope of the cue ‘might’ is ‘rain tomorrow’. Thus, the speculation cue is the word that expresses the speculation, while the words affected by the speculation are in the scope of that cue.
This task was the CoNLL-2010 Shared Task (BIBREF0), which had 3 different subtasks. Task 1B was speculation cue detection on the BioScope Corpus, Task 1W was weasel identification from Wikipedia articles, and Task 2 was speculation scope resolution from the BioScope Corpus. For each task, the participants were provided the train and test set, which is henceforth referred to as Task 1B CoNLL and Task 2 CoNLL throughout this paper.
For our experimentation, we use the sub corpora of the BioScope Corpus (BIBREF1), namely the BioScope Abstracts sub corpora, which is referred to as BA, and the BioScope Full Papers sub corpora, which is referred to as BF. We also use the SFU Review Corpus (BIBREF2), which is referred to as SFU.
This subtask of natural language processing, along with another similar subtask, negation detection and scope resolution, have been the subject of a body of work over the years. The approaches used to solve them have evolved from simple rule-based systems (BIBREF3) based on linguistic information extracted from the sentences, to modern deep-learning based methods. The Machine Learning techniques used varied from Maximum Entropy Classifiers (BIBREF4) to Support Vector Machines (BIBREF5,BIBREF6,BIBREF7,BIBREF8), while the deep learning approaches included Recursive Neural Networks (BIBREF9,BIBREF10), Convolutional Neural Networks (BIBREF11) and most recently transfer learning-based architectures like Bidirectional Encoder Representation from Transformers (BERT) (BIBREF12). Figures FIGREF1 and FIGREF1 contain a summary of the papers addressing speculation detection and scope resolution (BIBREF13, BIBREF5, BIBREF9, BIBREF3, BIBREF14, BIBREF15, BIBREF16, BIBREF17, BIBREF6, BIBREF11, BIBREF18, BIBREF10, BIBREF19, BIBREF7, BIBREF4, BIBREF8).
Inspired by the most recent approach of applying BERT to negation detection and scope resolution (BIBREF12), we take this approach one step further by performing a comparative analysis of three popular transformer-based architectures: BERT (BIBREF20), XLNet (BIBREF21) and RoBERTa (BIBREF22), applied to speculation detection and scope resolution. We evaluate the performance of each model across all datasets via the single dataset training approach, and report all scores including inter-dataset scores (i.e. train on one dataset, evaluate on another) to test the generalizability of the models. This approach outperforms all existing systems on the task of speculation detection and scope resolution. Further, we jointly train on multiple datasets and obtain improvements over the single dataset training approach on most datasets.
Contrary to results observed on benchmark GLUE tasks, we observe XLNet consistently outperforming RoBERTa. To confirm this observation, we apply these models to the negation detection and scope resolution task, and observe a continuity in this trend, reporting state-of-the-art results on three of four datasets on the negation scope resolution task.
The rest of the paper is organized as follows: In Section 2, we provide a detailed description of our methodology and elaborate on the experimentation details. In Section 3, we present our results and analysis on the speculation detection and scope resolution task, using the single dataset and the multiple dataset training approach. In Section 4, we show the results of applying XLNet and RoBERTa on negation detection and scope resolution and propose a few reasons to explain why XLNet performs better than RoBERTa. Finally, the future scope and conclusion is mentioned in Section 5.
we take this approach one step further by performing a comparative analysis of three popular transformer-based architectures: BERT, XLNet and RoBERTa, applied to speculation detection and scope resolution.
|
Where does the comparative analysis applied?
|
Speculation detection and scope resolution.
|
null | false
| null |
The King's Building is a Grade I listed building that forms part of the Strand Campus of King's College London in the United Kingdom. Originally named the College Building, the King's Building was designed by Sir Robert Smirke in the course of the College's foundation in 1829. As the founding building, it was built between 1829 and 1831 on land granted to King's College by the Government to complete the riverside frontage of Somerset House.
There are today a total of eight floors in the King's Building: Basement level, Ground level, Levels 1 to 4, Level 4U and Level 6. The King's Building houses a number of administrative departments (Estates & Facilities Offices, Admissions Office, Accommodation and Cashiers' Office), lecture theatres (most notably the Edmond J. Safra Lecture Theatre and Anatomy Lecture Theatre), the College Chapel, the Great Hall, various function rooms (Old Council Room, Committee Room, River Room, Somerset Room, St David's Room), study rooms and a catering outlet.
|
Extract a list of all the administrative departments and function rooms found in the King's Building based on this paragraph.
|
The administrative departments are Estates & Facilities Offices, Admissions Office, Accommodation and Cashiers' Office. The function rooms are Old Council Room, Committee Room, River Room, Somerset Room and St David's Room
|
1607.06275
| false
| null |
In order to train and evaluate open-domain factoid QA system for real-world questions, we build a new Chinese QA dataset named as WebQA. The dataset consists of tuples of (question, evidences, answer), which is similar to example in Figure FIGREF3 . All the questions, evidences and answers are collected from web. Table TABREF20 shows some statistics of the dataset.
In order to train and evaluate open-domain factoid QA system for real-world questions, we build a new Chinese QA dataset named as WebQA.
|
What languages do they experiment with?
|
The answers are shown as follows:
* Chinese
|
null | false
| 57
|
Electronic Healthcare records contain large volumes of unstructured data in different forms. Free text constitutes a large portion of such data, yet this source of richly detailed information often remains under-used in practice because of a lack of suitable methodologies to extract interpretable content in a timely manner. Here we apply network-theoretical tools to the analysis of free text in Hospital Patient Incident reports in the English National Health Service, to find clusters of reports in an unsupervised manner and at different levels of resolution based directly on the free text descriptions contained within them. To do so, we combine recently developed deep neural network text-embedding methodologies based on paragraph vectors with multi-scale Markov Stability community detection applied to a similarity graph of documents obtained from sparsified text vector similarities. We showcase the approach with the analysis of incident reports submitted in Imperial College Healthcare NHS Trust, London. The multiscale community structure reveals levels of meaning with different resolution in the topics of the dataset, as shown by relevant descriptive terms extracted from the groups of records, as well as by comparing a posteriori against hand-coded categories assigned by healthcare personnel. Our content communities exhibit good correspondence with well-defined hand-coded categories, yet our results also provide further medical detail in certain areas as well as revealing complementary descriptors of incidents beyond the external classification. We also discuss how the method can be used to monitor reports over time and across different healthcare providers, and to detect emerging trends that fall outside of pre-existing categories.
Electronic Healthcare records contain large volumes of unstructured data in different forms. Free text constitutes a large portion of such data, yet this source of richly detailed information often remains under-used in practice because of a lack of suitable methodologies to extract interpretable content in a timely manner.
|
Why does the paper take free text into the study?
|
Free text constitutes a large portion of such data, yet this source of richly detailed information often remains under-used in practice because of a lack of suitable methodologies to extract interpretable content in a timely manner.
|
null | false
| null |
What's the most popular tv show of all time in the USA?
|
Jeopardy!
|
|
null | false
| null |
Should I pursue a degree after graduating high school?
|
Whether or not you pursue a degree after graduating high school is an important but complex question that depends on a variety of factors. When assessing what the right choice is for you, it is important to think about the career you'd like to pursue. If you enjoy classwork and can afford a college education, that is one route which may provide opportunities to earn a higher salary later in life. However, if you prefer being active and would like to begin earning money immediately after graduating high school without accumulating significant debt, forgoing an advanced degree to begin work can be a great option. At the end of the day, pursuing something that you enjoy spending time on is the most important aspect of this decision and will be a better indicator of both satisfaction and success than whether or not you earn an advanced degree.
|
|
null | false
| 364
|
Automatic dialog/conversation systems have served humans for a long time in various fields, ranging from train routing nbcitetrain to museum guiding nbcitemuseum. In the above scenarios, the dialogs are domain-specific, and a typical approach to such in-domain systems is by human engineering, for example, using manually constructed ontologies nbciteyoungsigdial, natural language templates nbcitetemplate, and even predefined dialog states nbcitestatetracking.
Recently, researchers have paid increasing attention to open-domain, chatbot-style human-computer conversation, because of its important commercial applications, and because it tackles the real challenges of natural language understanding and generation nbciteretrieval1,acl,aaai. For open-domain dialogs, rules and temples would probably fail as we can hardly handle the great diversity of dialog topics and natural language sentences. With the increasing number of human-human conversation utterances available on the Internet, previous studies have developed data-oriented approaches in the open domain, which can be roughly categorized into two groups: retrieval systems and generative systems.
When a user issues an utterance (called a query), retrieval systems search for a most similar query in a massive database (which consists of large numbers of query-reply pairs), and respond to the user with the corresponding reply nbciteretrieval1,retrieval2. Through information retrieval, however, we cannot obtain new utterances, that is, all replies have to appear in the database. Also, the ranking of candidate replies is usually judged by surface forms (e.g., word overlaps, tf $\cdot $ idf features) and hardly addresses the real semantics of natural languages.
Generative dialog systems, on the other hand, can synthesize a new sentence as the reply by language models nbciteBoWdialog,acl,aaai. Typically, a recurrent neural network (RNN) captures the query's semantics with one or a few distributed, real-valued vectors (also known as embeddings); another RNN decodes the query embeddings to a reply. Deep neural networks allow complicated interaction by multiple non-linear transformations; RNNs are further suitable for modeling time-series data (e.g., a sequence of words) especially when enhanced with long short term memory (LSTM) or gated recurrent units (GRUs). Despite these, RNN also has its own weakness when applied to dialog systems: the generated sentence tends to be short, universal, and meaningless, for example, “I don't know” nbcitenaacl or “something” nbciteaaai. This is probably because chatbot-like dialogs are highly diversified and a query may not convey sufficient information for the reply. Even though such universal utterances may be suited in certain dialog context, they make users feel boring and lose interest, and thus are not desirable in real applications.
In this paper, we are curious if we can combine the above two streams of approaches for open-domain conversation. To this end, we propose an ensemble of retrieval and generative dialog systems. Given a user-issued query, we first obtain a candidate reply by information retrieval from a large database. The query, along with the candidate reply, is then fed to an utterance generator based on the “bi-sequence to sequence” (biseq2seq) model nbcitemultiseq2seq. Such sequence generator takes into consideration the information contained in not only the query but also the retrieved reply; hence, it alleviates the low-substance problem and can synthesize replies that are more meaningful. After that we use the scorer in the retrieval system again for post-reranking. This step can filter out less relevant retrieved replies or meaningless generated ones. The higher ranked candidate (either retrieved or generated) is returned to the user as the reply.
From the above process, we see that the retrieval and generative systems are integrated by two mechanisms: (1) The retrieved candidate is fed to the sequence generator to mitigate the “low-substance” problem; (2) The post-reranker can make better use of both the retrieved candidate and the generated utterance. In this sense, we call our overall approach an ensemble in this paper. To the best of our knowledge, we are the first to combine retrieval and generative models for open-domain conversation.
Experimental results show that our ensemble model consistently outperforms each single component in terms of several subjective and objective metrics, and that both retrieval and generative methods contribute an important portion to the overall approach. This also verifies the rationale for building model ensembles for dialog systems.
To the best of our knowledge, we are the first to combine retrieval and generative models for open-domain conversation.
|
Are they the first to combine retrieval and generative dialog systems?
|
Yes.
|
null | false
| null |
What is the show Parks and Rec about?
|
Parks and Rec is an American sitcom about the Parks and Recreation department of a small town in Indiana called Pawnee.
|
|
null | false
| 34
|
Since we focus on single-relation question answering in this work, our model decodes every question into a KB query that consists of exactly two elements–the topic entity, and the predicate. More formally, our model is a function $f(q, \lbrace e\rbrace , \lbrace p\rbrace )$ that takes as input a question $q$ , a set of candidate entities $\lbrace e\rbrace =e_1, ...,e_n$ , a set of candidate predicates $\lbrace p\rbrace =p_1,..., p_m$ , and produces a likelihood score $p(e_i, p_j|q)$ of generating entity $e_i$ and predicate $p_j$ given question $q$ for all $i\in {1...n}, j\in {1...m}$ .
As illustrated in Figure 1, our model consists of three components:
The details of each component are described in the following sections.
As illustrated in Figure 1, our model consists of three components:
|
How many components does the authors' model consist of?
|
Three.
|
2004.03925
| true
| null |
In this section, we present the details of the analysis performed on the data obtained pertaining to Twitter messages from January 2020 upto now, that is the time since the news of the Coronavirus outbreak in China was spread across nations. The word frequency data corresponding to the twitter messages has been taken from BIBREF16. The data source indicates that during March 11th to March 30th there were over 4 million tweets a day with the surge in the awareness. Also, the data prominently captures the tweets in English, Spanish, and French languages. A total of four datasets have been used to carry out the study.
In this section, we present the details of the analysis performed on the data obtained pertaining to Twitter messages from January 2020 upto now, that is the time since the news of the Coronavirus outbreak in China was spread across nations.
Also, the data prominently captures the tweets in English, Spanish, and French languages.
|
Do they collect only English data?
|
No.
|
null | false
| null |
What is the national bird of India
|
The Peacock
|
|
null | false
| null |
What is success
|
Satisfaction levels define the success, success defined for one person may not be true for other
|
|
null | false
| null |
What is a mortgage?
|
A mortgage loan or simply mortgage, in civil law jurisdicions known also as a hypothec loan, is a loan used either by purchasers of real property to raise funds to buy real estate, or by existing property owners to raise funds for any purpose while putting a lien on the property being mortgaged. The loan is "secured" on the borrower's property through a process known as mortgage origination. This means that a legal mechanism is put into place which allows the lender to take possession and sell the secured property ("foreclosure" or "repossession") to pay off the loan in the event the borrower defaults on the loan or otherwise fails to abide by its terms. The word mortgage is derived from a Law French term used in Britain in the Middle Ages meaning "death pledge" and refers to the pledge ending (dying) when either the obligation is fulfilled or the property is taken through foreclosure. A mortgage can also be described as "a borrower giving consideration in the form of a collateral for a benefit (loan)".
|
|
null | false
| null |
How do I calm down a toddler when they are having a tantrum?
|
Toddlers have big expectations and even bigger feelings and they do not yet hav the ability to identify and control their emotions and reactions; it's the parents job to help their toddlers learn to cope and adjust. Not all children are the same, so there is no "magic bullet" to calming down a child in a tantrum, but a couple things that can help when your child is having a melt-down:
1. Remember that their feelings are real and they are not trying to manipulate. Young children do not yet have the capacity to manipulate
2. Make sure the space is safe for the child, and any other children around.
3. Do not raise your voice to match their intensity! Use a calm, level tone with the child. Though this can be very difficult when you are also frustrated, a lower tone will help the child lower theirs as well.
3. Let the tantrum run out of steam.
4. Give a lot of hugs to your upset child, when they are ready
5. When your child is calmed down, talk about what happened! Give their feelings names (sad, angry, frustrated) to help your child identify the feeling when it happens next time, and talk with them about what they should do instead of throwing a tantrum (take a break, tell an adult/parent, deep breaths, etc.)
|
|
null | false
| 50
|
Equations are an important part of scientific articles, but many existing machine learning methods do not easily handle them. They are challenging to work with because each is unique or nearly unique; most equations occur only once. An automatic understanding of equations, however, would significantly benefit methods for analyzing scientific literature. Useful representations of equations can help draw connections between articles, improve retrieval of scientific texts, and help create tools for exploring and navigating scientific literature.
In this paper we propose equation embeddings (EqEmb), an unsupervised approach for learning distributed representations of equations. The idea is to treat the equation as a "singleton word," one that appears once but that appears in the context of other words. The surrounding text of the equation—and in particular, the distributed representations of that text—provides the data we need to develop a useful representation of the equation.
Figure FIGREF1 illustrates our approach. On the left is an article snippet BIBREF0 . Highlighted in orange is an equation; in this example it represents a neural network layer. We note that this particular equation (in this form and with this notation) only occurs once in the collection of articles (from arXiv). The representations of the surrounding text, however, provide a meaningful context for the equation. Those words allow us to learn its embedding, specifically as a "word" which appears in the context of its surroundings. The resulting representation, when compared to other equations' representations and word representations, helps find both related equations and related words. These are illustrated on the right.
EqEmbs build on exponential family embeddings BIBREF1 to include equations as singleton observations and to model equation elements such as variables, symbols and operators. Exponential family embeddings, like all embedding methods, define a context of each word. In our initial EqEmb, the context for the words is a small window, such as four or eight words, but the context of an equation is a larger window, such as sixteen words. Using these two types of contexts together finds meaningful representations of words and equations. In the next EqEmb, which builds on the first, we consider equations to be sentences consisting of equation units, i.e., variables, symbols, and operators. Equation units help model equations across two types of context—over the surrounding units and over the surrounding words.
We studied EqEmbs on four collections of scientific articles from the arXiv, covering four computer science domains: natural language processing (NLP), information retrieval (IR), artificial intelligence (AI) and machine learning (ML). We found that EqEmbs provide more efficient modeling than existing word embedding methods. We further carried out an exploratory analysis of a large set of INLINEFORM0 87k equations. We found that EqEmbs provide better models when compared to existing word embedding approaches. EqEmbs also provide coherent semantic representations of equations and can capture semantic similarity to other equations and to words.
In this paper we propose equation embeddings (EqEmb), an unsupervised approach for learning distributed representations of equations.
|
What kind of method the proposed equation embeddings are?
|
An unsupervised approach for learning distributed representations of equations.
|
null | false
| null |
What is Noether's theorem and how does it relate symmetries to conserved quantities?
|
Emmy Noether proposed a theorem that states that for every symmetry in physics, there is a conserved quantity. For example, for the symmetry that the laws of physics are unchanged by a translation in space, a rotation in space, or an offset in time, the corresponding conserved quantities are linear momentum, angular momentum, and the conservation of energy.
|
|
null | false
| null |
Who is Mauricio Macri?
|
Mauricio Macri, born 8 February 1959, is an Argentine businessman and politician who served as the President of Argentina from 2015 to 2019. He has been the leader of the Republican Proposal (PRO) party since its founding in 2005. He previously served as 5th Chief of Government of Buenos Aires from 2007 to 2015, and was a member of the Chamber of Deputies representing Buenos Aires from 2005 to 2007. Ideologically, he identifies himself as a liberal and conservative on the Argentine centre-right.
|
|
null | false
| null |
Which is a species of fish? Pearl or Pearlfish
|
Pearlfish
|
|
null | false
| null |
Larry Joe Bird (born December 7, 1956) is an American former professional basketball player, coach, and executive in the National Basketball Association (NBA). Nicknamed "the Hick from French Lick" and "Larry Legend", Bird is widely regarded as one of the greatest basketball players of all time. He is the only person in NBA history to be named Rookie of the Year, Most Valuable Player, Finals MVP, All-Star MVP, Coach of the Year, and Executive of the Year. Growing up in French Lick, Indiana, he was a local basketball star. Highly recruited, he initially signed to play college basketball for coach Bob Knight of the Indiana Hoosiers, but Bird dropped out after one month and returned to French Lick to attend a local community college. The next year he attended the smaller Indiana State University, ultimately playing three years for the Sycamores. Drafted by the Boston Celtics with the sixth overall pick in the 1978 NBA draft after his second year at Indiana State, Bird elected to stay in college and returned for the 1978–1979 season. He then led his team to an undefeated regular season. The season finished with a national championship game match-up of Indiana State against Michigan State and featured a highly anticipated match-up of Bird against Michigan State great Magic Johnson, thus beginning a career-long rivalry that the two shared for over a decade.
https://en.wikipedia.org/wiki/Larry_Bird
|
Given the reference text below about Larry Bird, tell me where he was born, what he did for a living, and who his greatest professional rival was.
|
Larry Bird was born in French Lick. He played and coached basketball for a living, and his main rival was Magic Johnson.
|
null | false
| null |
Glowworm, was briefly the name of a coastal destroyer launched on 12 December 1906 and renamed HMS TB7. It was sold in May 1921.
HMS Glowworm (1916), an Insect-class gunboat, launched on 5 February 1916 and sold in 1928.
HMS Glowworm (H92), a G-class destroyer launched on 22 July 1935, sunk on 8 April 1940 by the German heavy cruiser Admiral Hipper off Norway.
Glowworm was allocated to a G-class destroyer under construction at the William Denny shipyard at Dumbarton in 1945. The vessel was originally called HMS Guinevere but was renamed in September 1945 to HMS Glowworm, and renamed again in October to HMS Gift. Construction was cancelled on 1 December 1945 before completion.
|
When was HMS Glowworm (H92), a G-class destroyer launched?
|
The HMS Glowworm (H92), a G-class was launched in July 22nd, 1935.
|
null | false
| 366
|
Modeling temporal and sequential data, which is crucial in machine learning, can be applied in many areas, such as speech and natural language processing. Deep neural networks (DNNs) have garnered interest from many researchers after being successfully applied in image classification BIBREF0 and speech recognition BIBREF1 . Another type of neural network, called a recurrent neural network (RNN) is also widely used for speech recognition BIBREF2 , machine translation BIBREF3 , BIBREF4 and language modeling BIBREF5 , BIBREF6 . RNNs have achieved many state-of-the-art results. Compared to DNNs, they have extra parameters for modeling the relationships of previous or future hidden states with current input where the RNN parameters are shared for each input time-step.
Generally, RNNs can be separated by a simple RNN without gating units, such as the Elman RNN BIBREF7 , the Jordan RNN BIBREF8 , and such advanced RNNs with gating units as the Long-Short Term Memory (LSTM) RNN BIBREF9 and the Gated Recurrent Unit (GRU) RNN BIBREF4 . A simple RNN usually adequate to model some dataset and a task with short-term dependencies like slot filling for spoken language understanding BIBREF10 . However, for more difficult tasks like language modeling and machine translation where most predictions need longer information and a historical context from each sentence, gating units are needed to achieve good performance. With gating units for blocking and passing information from previous or future hidden layer, we can learn long-term information and recursively backpropagate the error from our prediction without suffering from vanishing or exploding gradient problems BIBREF9 . In spite of this situation, the concept of gating mechanism does not provide an RNN with a more powerful way to model the relation between the current input and previous hidden layer representations.
Most interactions inside RNNs between current input and previous (or future) hidden states are represented using linear projection and addition and are transformed by the nonlinear activation function. The transition is shallow because no intermediate hidden layers exist for projecting the hidden states BIBREF11 . To get a more powerful representation on the hidden layer, Pascanu et al. BIBREF11 modified RNNs with an additional nonlinear layer from input to the hidden layer transition, hidden to hidden layer transition and also hidden to output layer transition. Socher et al. BIBREF12 , BIBREF13 proposed another approach using a tensor product for calculating output vectors given two input vectors. They modified a Recursive Neural Network (RecNN) to overcome those limitations using more direct interaction between two input layers. This architecture is called a Recursive Neural Tensor Network (RecNTN), which uses a tensor product between child input vectors to represent the parent vector representation. By adding the tensor product operation to calculate their parent vector, RecNTN significantly improves the performance of sentiment analysis and reasoning on entity relations tasks compared to standard RecNN architecture. However, those models struggle to learn long-term dependencies because the do not utilize the concept of gating mechanism.
In this paper, we proposed a new RNN architecture that combine the gating mechanism and tensor product concepts to incorporate both advantages in a single architecture. Using the concept of such gating mechanisms as LSTMRNN and GRURNN, our proposed architecture can learn temporal and sequential data with longer dependencies between each input time-step than simple RNNs without gating units and combine the gating units with tensor products to represent the hidden layer with more powerful operation and direct interaction. Hidden states are generated by the interaction between current input and previous (or future) hidden states using a tensor product and a non-linear activation function allows more expressive model representation. We describe two different models based on LSTMRNN and GRURNN. LSTMRNTN is our proposed model for the combination between a LSTM unit with a tensor product inside its cell equation and GRURNTN is our name for a GRU unit with a tensor product inside its candidate hidden layer equation.
In Section "Background" , we provide some background information related to our research. In Section "Proposed Architecture" , we describe our proposed RNN architecture in detail. We evaluate our proposed RNN architecture on word-level and character-level language modeling tasks and reported the result in Section "Experiment Settings" . We present related works in Section "Related Work" . Section "Conclusion" summarizes our paper and provides some possible future improvements.
In this paper, we proposed a new RNN architecture that combine the gating mechanism and tensor product concepts to incorporate both advantages in a single architecture.
|
What does the architecture combine?
|
Combine the gating mechanism and tensor product concepts.
|
null | false
| null |
What is the best way to put a toddler to sleep at night?
|
Putting a toddler to sleep can be one of the most difficult things. The best way is to make sure that the toddler is very active during the day and is tired enough to fall asleep naturally. The toddler should get enough sunlight. Avoid or minimize exposure to digital screens such as television, tablet, and smartphones. Create a daily bedtime routine. For example, you can start the routine by feeding a bottle of milk, reading a story from a picture book, singing a lullaby, and shushing. For some babies and toddlers white noise helps to fall asleep.
|
|
null | false
| 490
|
Background Deep neural networks (DNNs), in particular convolutional neural networks (CNNs), have been powerful tools for solving various image inverse problems such as denoising, inpainting, and super resolution. Conventional wisdom believes that is owing to DNNs' universal approximation ability and learning from massive training data. Yet, recent studies have revealed the specific architectures of CNNs have the inductive bias to represent and generate natural images well, and such favorable architecture inductive bias can work independently from fitting specific training sets.
For example, deep image prior (DIP) shows an untrained neural network can be used as a handcrafted prior that transfers well across multiple inverse problems. The authors attributed the success to the CNN architecture itself, that appeared to possess high noise impedance even with only random initializations. As another example, in compressive sensing,; replaced the common structural assumptions such as sparsity with a pretrained generative adversarial networks. The underlying rationale lies in that a pre-trained generator should (approximately) represent the notion of a vector being or more likely in the target domain such as natural images; in other words, a sample more like a natural image will be closer to the output range of the pre-trained generator. In this paper, we refer to such general-purpose image prior parameterized by (either untrained or pre-trained) DNNs as DNN-based image priors.
Recall that, classical image regularizers in the spatial or frequency domains are often not learningbased, or rely on compact learning models. On contrary, DNN-based image priors have a massive number of parameters (we compare the parameter numbers between the full model and the sparse subnetwork in Table 2), typically magnitudes more than the image (even the image size) size. The two extremes invite the natural question: Can we identify highly compact DNN-based image priors, that are same effective? We note that, diving into this question has twofold appeals. On the algorithmic side, that could help us understand further how the topology and connectivity of CNN architecture itself will affect the effectiveness of those priors, and to what extent sparsity could be relevant. On the practical side, if we could provide an affirmative answer to this question, that would potentially lead to more computationally savings when applying those DNN-based priors in practice, leading to faster restoration or computational imaging with them.
Towards the above question, the tool we refer to in this paper is the recently emerged Lottery Ticket Hypothesis (LTH). LTH suggests that every dense DNN has an extremely sparse "matching subnetwork", that can be trained in isolation to match the original dense DNN's accuracy. While the vanilla LTH studies training from random scratch, the latest works also extend similar findings to fine-tuning the pre-trained models. LTH has widespread success in image classification, language modeling, reinforcement learning and multi-modal learning, e.g.,.
Our Contributions Drawing inspirations from the LTH literature, we conjecture and empirically study a novel "lottery image prior" (LIP), stated as: Given an (untrained or trained) DNN-based image prior, it will have a sparse subnetwork that can be trained in isolation, to match the original DNN's performance when being applied as a prior to regularizing various image inverse problems.
Studying this new problem is, however, NOT a naive extension from the existing LTH methods, owing to several technical barriers: (a) till now, LTH has not been demonstrated for image inverse problem or DNN-based priors, to our best knowledge. Most LTH works studied discriminative tasks, with one exception. It is therefore uncertain whether high-sparsity DNN is still viable for reconstruction-oriented tasks; (b) existing LTH works typically require a full training set to locate the sparse subnetwork mask, whereas our LIP settings are only not data-rich. For example, DIP needs the DNN to be trained to overfit one specific image, making it drastically different from previous problems; (c) the objectives between finding the sparse mask (e.g., learning the prior) and fitting the sparse subnetwork (e.g., using the prior) are often unaligned in LIP problems. For example, DIP will overfit a corrupted input image (during which the sparse mask will be found) in order to reconstruct a clean output image (when the found sparse subnetwork will be used); the pre-trained generator will also be used towards a different goal (regularizing compressive sensing) from their original pre-training task (generating realistic images).
Our extensive experimental study confirms the existence of LIP in two representative settings: (i) image restoration with the deep image prior, using an untrained DNN (as shown in Fig.); and
(ii) compressive sensing image reconstruction, using a pre-trained GAN generator. Using iterative magnitude pruning (IMP) with surrogate tasks (the overview of our work paradigm is in Fig.), we can successfully locate the LIP subnetworks at the sparsity range of 20%-86.58% in setting i; and those at the sparsity range of 5%-36% in setting ii. Those LIP subnetworks also possess high transferability. For example, the LIP ticket found in the setting i transfer well across not only different images, but also different tasks such as denoising, inpainting and super-resolution. Our contributions are summarized below:
• The first comprehensive study on LTH in DNN-based image priors and inverse problems, establishing the "lottery image prior" (LIP) and demonstrating the prevailing relevance of LTH more broadly than previously typical settings. The last column (in blue) intends to display the results with the most extremely sparse subnetwork.
• The investigation of both untrained and pre-trained DNNs as image priors, verifying that LIP subnetworks can be found in both settings, by overcoming several impediments such as severely limited training data and task objective mismatch. • High transferability of LIP subnetworks across data/datasets, and various inverse problem tasks. Our finding reflects the underlying common image prior that is agnostic to specific data or task, through the lens of the CNN architecture together with sparsity.
Figure 11: Layer-wise sparsity ratio results of LIP, SNIP and randomly pruned tickets. Note that we summarize the sparsity ratio of each layer: the ratio of the number of parameters whose values are equal to zero to the number of total parameters of the layer. And the x-axis of these figures is composed of the serial numbers of model layers. We sampled subnetworks with four different sparsities (sparsity = 36%, 59%, 89%, 95%) to observe.****Figure 13: Learning curves using four different training targets: a clean image (Baby.png), the same image added with noises, the same randomly scrambled and white noise. Note that we use four different models: the LIP subnetwork (S = 89%), randomly pruned subnetwork (S = 89%), SNIP subnetwork (S = 89%) and the dense model (S = 0%). And we trained them in isolation in the same experimental settings for 10000 iterations.****Figure 2: LIP visual results: inpainting (row 1), super-resolution (rows 2/3) and denoising (row 4). The last column (in blue) intends to display the results with the most extremely sparse subnetwork.****To better explain the success of the LIP subnetworks, inspired by the Figure 2 of (Ulyanov et al., 2018), we further train the obtained subnetworks (LIP, SNIP and random pruning) with four different targets: 1) a natural image, 2) the same added with noise, 3) the same after randomly permuting the pixels and 4) the white noise.****Interestingly, we also find that SNIP subnetworks perform similarly with the dense model. Meanwhile, the parametrization of LIP subnetworks offers the higher impedance to noise and the lower impedance to signal than the dense model, which indicates that the separation of high frequency and low frequency is more obvious for winning network architectures.
|
Why the specific lottery image prior structure is superior to other structures in DIP tasks?
|
Thanks for the advice. We add the experiments of studying the layer-wise sparsity ratio and the learning curves of different subnetworks, to add to the interpretability of what is special about LIP subnetworks. The results are summarized in Figure 11 and 13 in the section - Supplementary Materials of the paper. As shown in Figure 11, the structure of the LIP subnetwork is drastically different from those found by SNIP and random pruning, in particular the distribution of layerwise sparsity ratios. LIP tends to preserve weights of the earlier layers (closer to the input), while pruning the latter layers more aggressively (e.g, Figure 11(a).). In contrast, SNIP tends to prune much more of the earlier layers compared to the latter ones. Random pruning by default prunes each layer at approximately the same ratio. Comparing the three methods seems to suggest that for finding effective and transferable LIP subnetworks, specifically keeping more weights at earlier layers more is important. This is an explainable finding because, in image restoration tasks, the low-level features (color, texture, shape, etc.) presumably matter more and are more transferable, than the high-level features (object categories, etc.). The earlier layers are known to capture more low-level image features, hence contributing more to retraining the image restoration performance with DIP. To better explain the success of the LIP subnetworks, inspired by the Figure 2 of the paper “Deep Image Prior”, we further train the obtained subnetworks (LIP, SNIP and random pruning) with four different targets: 1) a natural image, 2) the same added with noise, 3) the same after randomly permuting the pixels and 4) the white noise. We also train the dense model as the baseline. The experimental results are summarized in Figure 13. We first observe that for LIP, SNIP and dense models, the optimization converges much faster in case 1) and 2) than in case 3) and 4). But the randomly pruned subnetworks have failed in all cases. Interestingly, we also find that SNIP subnetworks perform similarly with the dense model. Meanwhile, the parametrization of LIP subnetworks offers a higher impedance to noise and a lower impedance to signal than the dense model, which indicates that the separation of high frequency and low frequency is more obvious for winning network architectures.
|
null | false
| null |
I'm confused as to which of these food staples have gluten and which are gluten-free. From this list can you tell me which is which: Pasta, rice, potatoes, noodles and bread.
|
Rice and potatoes do not contain gluten, so they are gluten-free. However, pasta, noodles and bread usually contain gluten.
There are less common varieties of these staples that do not contain gluten, but you will need to check the packaging details to ensure that they are gluten free.
|
|
null | false
| null |
Does the position of a dog's tail communicate how it is feeling?
|
Yes, the position of a dog's tail will communicate how the dog is feeling. For example, if a dog's tail wags back and forth, the dog feels relaxed. If a dog's tail is standing straight up, the dog is alert and tense. When a dog's tail is tucked down low, it is nervous. If the dog's tail is moving in circles, the dog is happy.
|
|
null | false
| null |
Give me a list of places to go in San Diego, CA with kids
|
1. San Diego Zoo
2. Sea World
3. Legoland
4. Balboa Park
5. Belmont Park
6. Any of the beaches
|
|
null | false
| null |
What other names are used to refer to Taweret, the Egyptian goddess:
|
The Egyptian goddess Taweret can be called Taouris, Taueret, Taurt, Tawaret, Ta-weret, Thoeris, Thouéris, Toeris, Tuart, Tuat or Twert.
|
|
null | false
| null |
He was named bishop of Saint-Omer on 29 June 1775, then bishop of Carcassonne in 1778. In 1788, he became the Archbishop of Bourges. A deputy to Estates-General of 1789, on the French Revolution he emigrated to Wolfenbüttel, where he lived with the archbishop of Rheims, Talleyrand-Périgord. The 1801 Concordat between France and the Pope forced him to resign, but allowed him to return to Rabastens, where he then lived until his death.
|
Where did the bishop live after 1801?
|
The bishop lived in Rabastens after 1801, where he resided until his death, after being forced to resign by the pope.
|
null | false
| null |
What is gravel biking?
|
Gravel biking is riding mainly on unpaved roads, opening up large amounts of terrain, a standard road bike normally can't access. This is due to the tires being wider, with more volume and the frame geometry being more compliant and suited to rougher terrain. Gravel biking is the fastest growing cycling genre globally and has large high profile race events.
|
|
null | false
| 186
|
Story-telling is on the frontier of current text generation technology: stories must remain thematically consistent across the complete document, requiring modeling very long range dependencies; stories require creativity; and stories need a high level plot, necessitating planning ahead rather than word-by-word generation BIBREF0 .
We tackle the challenges of story-telling with a hierarchical model, which first generates a sentence called the prompt describing the topic for the story, and then conditions on this prompt when generating the story. Conditioning on the prompt or premise makes it easier to generate consistent stories because they provide grounding for the overall plot. It also reduces the tendency of standard sequence models to drift off topic.
We find that standard sequence-to-sequence (seq2seq) models BIBREF1 applied to hierarchical story generation are prone to degenerating into language models that pay little attention to the writing prompt (a problem that has been noted in other domains, such as dialogue response generation BIBREF2 ). This failure is due to the complex and underspecified dependencies between the prompt and the story, which are much harder to model than the closer dependencies required for language modeling (for example, consider the subtle relationship between the first sentence and prompt in Figure FIGREF1 ).
To improve the relevance of the generated story to its prompt, we introduce a fusion mechanism BIBREF3 where our model is trained on top of an pre-trained seq2seq model. To improve over the pre-trained model, the second model must focus on the link between the prompt and the story. For the first time, we show that fusion mechanisms can help seq2seq models build dependencies between their input and output.
Another major challenge in story generation is the inefficiency of modeling long documents with standard recurrent architectures—stories contain 734 words on average in our dataset. We improve efficiency using a convolutional architecture, allowing whole stories to be encoded in parallel. Existing convolutional architectures only encode a bounded amount of context BIBREF4 , so we introduce a novel gated self-attention mechanism that allows the model to condition on its previous outputs at different time-scales.
To train our models, we gathered a large dataset of 303,358 human generated stories paired with writing prompts from an online forum. Evaluating free form text is challenging, so we also introduce new evaluation metrics which isolate different aspects of story generation.
Experiments show that our fusion and self-attention mechanisms improve over existing techniques on both automated and human evaluation measures. Our new dataset and neural architectures allow for models which can creatively generate longer, more consistent and more fluent passages of text. Human judges prefer our hierarchical model's stories twice as often as those of a non-hierarchical baseline.
Existing convolutional architectures only encode a bounded amount of context, so we introduce a novel gated self-attention mechanism that allows the model to condition on its previous outputs at different time-scales.
|
What's the function of the novel gated self-attention mechanism?
|
It allows the authors' model to condition on its previous outputs at different time-scales.
|
null | false
| null |
The game is set 200 years after the events of Oblivion, and takes place in Skyrim, the northernmost province of Tamriel. Its main story focuses on the player's character, the Dragonborn, on their quest to defeat Alduin the World-Eater, a dragon who is prophesied to destroy the world. Over the course of the game, the player completes quests and develops the character by improving skills. The game continues the open world tradition of its predecessors by allowing the player to travel anywhere in the game world at any time, and to ignore or postpone the main storyline indefinitely.
|
What is Skyrim?
|
Skyrim is a medieval fantasy game based in the fictional world of Tamriel. With it being the fifth game in the Elder Scrolls franchise, it has plenty of lore to draw upon. Like its predecessors, the game is an open world RPG with plenty of quests to do, monsters to kill, items to loot, skills to level and weapons to craft. The map and setting of Skyrim is heavily influenced by Norse mythology. It has become one of the most famous games to be created by Bethesda.
|
null | false
| null |
When did Delaware join the United States?
|
Delaware was admitted to the union on December 7, 1787.
|
|
null | false
| 335
|
While the demand for physical and manual labor is gradually declining, there is a growing need for a workforce with soft skills. Which soft skill do you think would be the most valuable in your daily life? According to an article in Forbes BIBREF0 , 70% of employed Americans agree that public speaking skills are critical to their success at work. Yet, it is one of the most dreaded acts. Many people rate the fear of public speaking even higher than the fear of death BIBREF1 . To alleviate the situation, several automated systems are now available that can quantify behavioral data for participants to reflect on BIBREF2 . Predicting the viewers' ratings from the speech transcripts would enable these systems to generate feedback on the potential audience behavior.
Predicting human behavior, however, is challenging due to its huge variability and the way the variables interact with each other. Running Randomized Control Trials (RCT) to decouple each variable is not always feasible and also expensive. It is possible to collect a large amount of observational data due to the advent of content sharing platforms such as YouTube, Massive Open Online Courses (MOOC), or ted.com. However, the uncontrolled variables in the observational dataset always keep a possibility of incorporating the effects of the “data bias” into the prediction model. Recently, the problems of using biased datasets are becoming apparent. BIBREF3 showed that the error rates in the commercial face-detectors for the dark-skinned females are 43 times higher than the light-skinned males due to the bias in the training dataset. The unfortunate incident of Google's photo app tagging African-American people as “Gorilla” BIBREF4 also highlights the severity of this issue.
We address the data bias issue as much as possible by carefully analyzing the relationships of different variables in the data generating process. We use a Causal Diagram BIBREF5 , BIBREF6 to analyze and remove the effects of the data bias (e.g., the speakers' reputations, popularity gained by publicity, etc.) in our prediction model. In order to make the prediction model less biased to the speakers' race and gender, we confine our analysis to the transcripts only. Besides, we normalize the ratings to remove the effects of the unwanted variables such as the speakers' reputations, publicity, contemporary hot topics, etc.
For our analysis, we curate an observational dataset of public speech transcripts and other meta-data collected from the ted.com website. This website contains a large collection of high-quality public speeches that are freely available to watch, share, rate, and comment on. Every day, numerous people watch and annotate their perceptions about the talks. Our dataset contains 2231 public speech transcripts and over 5 million ratings from the spontaneous viewers of the talks. The viewers annotate each talk by 14 different labels—Beautiful, Confusing, Courageous, Fascinating, Funny, Informative, Ingenious, Inspiring, Jaw-Dropping, Long-winded, Obnoxious, OK, Persuasive, and Unconvincing.
We use two neural network architectures in the prediction task. In the first architecture, we use LSTM BIBREF7 for a sequential input of the words within the sentences of the transcripts. In the second architecture, we use TreeLSTM BIBREF8 to represent the input sentences in the form of a dependency tree. Our experiments show that the dependency tree-based model can predict the TED talk ratings with slightly higher performance (average F-score 0.77) than the word sequence model (average F-score 0.76). To the best of our knowledge, this is the best performance in the literature on predicting the TED talk ratings. We compare the performances of these two models with a baseline of classical machine learning techniques using hand-engineered features. We find that the neural networks largely outperform the classical methods. We believe this gain in performance is achieved by the networks' ability to capture better the natural relationship of the words (as compared to the hand engineered feature selection approach in the baseline methods) and the correlations among different rating labels.
We use a Causal Diagram to analyze and remove the effects of the data bias (e.g., the speakers’ reputations, popularity gained by publicity, etc.) in our prediction model.
|
What effects in the prediction of public speech transcripts did the authors hope to analyze and remove?
|
The authors want to analyze and remove the effects of the data bias in the prediction of public speech transcripts.
|
null | false
| null |
Give me the list of Greek goddesses starting by 'B':
|
Here is the list of Greek goddesses starting by 'B':
- Baubo
- Benthesikyme
- Bia
- Borysthenis
- Brimo
- Brizo
- Bronte
|
|
null | false
| null |
Why capri is so famous
|
Capri is famous Italian island for it's nice ocean color and it's mountain view
|
|
null | false
| null |
Which country produces the most wine on an annual basis?
|
Italy
|
|
null | false
| null |
What is the fastest production car in the world?
|
The fastest production car in the world is currently the Bugatti Chiron Supersport 300+ due to its actually demonstrated 304mph top speed.
|
|
null | false
| null |
Who directed the first two episodes of season three of Game of Thrones?
|
Daniel Minahan directed "Valar Dohaeris" and "Dark Wings, Dark Words" which are the first two episodes of season three of Game of Thrones.
|
|
null | false
| null |
What are the qualities of a good manager?
|
Qualities that make a good manager could vary. Different managers bring different things to the table in different situations to help the team and company. Here are a few qualities that are generally useful:
1. Hiring: Primary job for a manager is to build a good team which is not possible without hiring the team.
2. Building trust: It's important for a manager to build trust with each individual on their team but also within the team members themselves to create a culture for collaboration
3. Problem solving: Managers must be able to recognize problems and help the team find resolutions either by directly helping them or finding the right help for the team
4. Prioritization: Teams often have more work than they have people and managers play a critical role in prioritizing tasks to ensure the higher impact tasks are done first.
5. Flexibility: Managers must be willing to play different roles as the situation demands it ranging from hands on domain expertise to conflict resolution to culture builder.
|
|
null | false
| null |
What is an apple?
|
An apple is an edible fruit produced by an apple tree. Apple trees are cultivated worldwide and are the most widely grown species in the genus Malus.
|
|
null | false
| null |
The climate of Andhra Pradesh varies considerably, depending on the geographical region. Summers last from March to June. In the coastal plain, the summer temperatures are generally higher than the rest of the state, with temperature ranging between 20 and 41 °C (68 and 106 °F). July to September is the season for tropical rains. About one-third of the total rainfall is brought by the northeast monsoon. October and November see low-pressure systems and tropical cyclones form in the Bay of Bengal which, along with the northeast monsoon, bring rains to the southern and coastal regions of the state.
November, December, January, and February are the winter months in Andhra Pradesh. Since the state has a long coastal belt the winters are not very cold. The range of winter temperature is generally 12 to 30 °C (54 to 86 °F). Lambasingi in Visakhapatnam district is also nicknamed as the "Kashmir of Andhra Pradesh" due to its relatively cool climate as compared to others and the temperature ranges from 0 to 10 °C (32 to 50 °F).
|
From the passage provided, extract the winter months in Andhra Pradesh. Separate them with a comma.
|
November, December, January, February
|
null | false
| null |
Thor (from Old Norse: Þórr) is a prominent god in Germanic paganism. In Norse mythology, he is a hammer-wielding god associated with lightning, thunder, storms, sacred groves and trees, strength, the protection of humankind, hallowing, and fertility. Besides Old Norse Þórr, the deity occurs in Old English as Þunor, in Old Frisian as Thuner, in Old Saxon as Thunar, and in Old High German as Donar, all ultimately stemming from the Proto-Germanic theonym *Þun(a)raz, meaning 'Thunder'.
|
Who is Thor?
|
Thor is the god of thunder in Norse mythology. He wields a powerful hammer that only he can pick up. Thor has appeared as a character in Marvel comic books since 1962 and in at least 10 movies.
|
null | false
| null |
Neeraj Chopra the Tokyo Olympic Gold Medallist won the Silver Medal at
|
Diamond League Meet in Stockholm
|
|
null | false
| 234
|
The ability to continuously learn and accumulate knowledge throughout a lifetime and reuse it effectively to adapt to a new problem quickly is a hallmark of general intelligence. State-of-the-art machine learning models work well on a single dataset given enough training examples, but they often fail to isolate and reuse previously acquired knowledge when the data distribution shifts (e.g., when presented with a new dataset)—a phenomenon known as catastrophic forgetting BIBREF0 , BIBREF1 .
The three main approaches to address catastrophic forgetting are based on: (i) augmenting the loss function that is being minimized during training with extra terms (e.g., a regularization term, an optimization constraint) to prevent model parameters learned on a new dataset from significantly deviating from parameters learned on previously seen datasets BIBREF2 , BIBREF3 , BIBREF4 , (ii) adding extra learning phases such as a knowledge distillation phase, an experience replay BIBREF5 , BIBREF6 , and (iii) augmenting the model with an episodic memory module BIBREF7 . Recent methods have shown that these approaches can be combined—e.g., by defining optimization constraints using samples from the episodic memory BIBREF8 , BIBREF9 .
In language learning, progress in unsupervised pretraining BIBREF10 , BIBREF11 , BIBREF12 has driven advances in many language understanding tasks BIBREF13 , BIBREF14 . However, these models have been shown to require a lot of in-domain training examples, rapidly overfit to particular datasets, and are prone to catastrophic forgetting BIBREF15 , making them unsuitable as a model of general linguistic intelligence.
In this paper, we investigate the role of episodic memory for learning a model of language in a lifelong setup. We propose to use such a component for sparse experience replay and local adaptation to allow the model to continually learn from examples drawn from different data distributions. In experience replay, we randomly select examples from memory to retrain on. Our model only performs experience replay very sparsely to consolidate newly acquired knowledge with existing knowledge in the memory into the model. We show that a 1% experience replay to learning new examples ratio is sufficient. Such a process bears some similarity to memory consolidation in human learning BIBREF16 . In local adaptation, we follow Memory-based Parameter Adaptation BIBREF7 and use examples retrieved from memory to update model parameters used to make a prediction of a particular test example.
Our setup is different than a typical lifelong learning setup. We assume that the model only makes one pass over the training examples, similar to BIBREF9 . However, we also assume neither our training nor test examples have dataset identifying information (e.g., a dataset identity, a dataset descriptor). Our experiments focus on lifelong language learning on two tasks—text classification and question answering. BIBREF17 show that many language processing tasks (e.g., classification, summarization, natural language inference, etc.) can be formulated as a question answering problem. We argue that our lifelong language learning setup—where a model is presented with question-answer examples without an explicit identifier about which dataset (distribution) the examples come from—is a more realistic setup to learn a general linguistic intelligence model.
Our main contributions in this paper are:
In this paper, we investigate the role of episodic memory for learning a model of language in a lifelong setup.
|
What does the paper investigate?
|
The role of episodic memory for learning a model of language in a continual (lifelong) learning setup.
|
null | false
| null |
Where should I live in Indiana?
|
There are many great places to live in Indiana. It depends on what is most important to you. If living near a body of water is important to you, you may want to look in northern Indiana near many lakes and Lake Michigan. If great schools are important to you, you might want to look at Zionsville, Indiana. This city has the best public school system in the state. Lastly, if you want to live near a college campus, you might enjoy a town like Bloomington or West Lafayette to be near Indiana University or Purdue University, respectively.
|
|
null | false
| 410
|
Figure 4: Training curves on D4RL continuous control benchmark across five random seeds. The
shaded areas represent the standard deviation across different seeds.
|
How many seeds were used and what are do the shaded areas represent?
|
As indicated by the caption of Figure 4, we used five random seeds, and the shared areas represent the standard deviation across different seeds.
|
null | false
| null |
Where can I mine diamonds in the United States?
|
The only place you can mine diamonds in the United States is located in Murfreesboro, Arkansas at the Crater of Diamonds State Park. It's open to the public and you get to keep whatever you find!
|
|
null | false
| null |
London is a 2005 Indian Tamil-language comedy-drama film written by Badri and directed by Sundar C. Bhaskaran hides the money before getting arrested, and the place is known only to his lawyer Vedimuthu. Aishwarya pretends to be in love with Vedimuthu to find where the money is hidden.
|
In which movie or plot the character vedimuthu was established
|
Vedimuthu character was potrayed by Vadivelu and become famous for his comic potrayal of act of love with mistress Aishwarya (potrayed by mumtaz) and getting caught in the act to his wife Baby(potrayed by Nalini).
|
null | false
| null |
What is Agent Fox Mulder's badge number?
|
JTT047101111
|
|
null | false
| null |
Who was offered the presidency of Israel in 1952 (turned down)
|
Albert Einstein
|
|
null | false
| null |
What is VCF?
|
Variant Call Format (VCF) is a file format used for storing and transferring genetic data. It is used for storing and comparing genetic data, and for representing genetic variants. VCF is a binary format that stores genetic data, including sequence data, variant data, and annotation data.
|
|
null | false
| null |
What are some of the easiest house plants to take care of?
|
Snake plant, Pothos Plant, ZZ Plant, Philodendron, Succulents
|
|
null | false
| null |
The U.S. Space Force traces its roots to the beginning of the Cold War, with the first Army Air Forces space programs starting in 1945. In 1954, the Western Development Division, under General Bernard Schriever, was established as the first dedicated space organization within the U.S. Armed Forces and continues to exist as the Space Force's Space and Missile Systems Center. Military space forces were organized under several different Air Force major commands until they were unified when Air Force Space Command was established on 1 September 1982. U.S. space forces first began conducting combat support operations in the Vietnam War and continued to provide satellite communications, weather, and navigation support during the 1982 Falklands War, 1983 United States invasion of Grenada, 1986 United States bombing of Libya, and 1989 United States invasion of Panama. The first major employment of space forces culminated in the Gulf War, where they proved so critical to the U.S.-led coalition that it is sometimes referred to as the first "space war".
The first discussions of creating a military space service occurred in 1958, and it was nearly established in 1982 by President Ronald Reagan as part of the Strategic Defense Initiative, but the idea dissolved with treaty compliance concerns. A 2001 Space Commission argued for the creation of a Space Corps around 2007–2011, but no action was taken by the Obama Administration. On 20 December 2019, the United States Space Force Act was signed into law as part of the National Defense Authorization Act, reorganizing Air Force Space Command and other space elements into the United States Space Force, and creating the first new independent military service since the Army Air Forces were reorganized as the U.S. Air Force in 1947.
|
From the passage below, please extract the official starting date of the United States Space Force.
|
The United States Space Force Act was signed into law as part of the National Defense Authorization Act on December 20, 2019. This act reorganized the Air Force Space Command and other space elements into the United States Space Force, and officially created the first new independent military service since the Army Air Forces were reorganized as the U.S. Air Force in 1947.
|
null | false
| null |
How do you plan a trip?
|
There are many different ways to plan a trip. In addition, there are many different factors to include in order to plan a trip. First of all, it is important to know the type of travel. Is this a business trip? Is this family trip?
Next you need to decide on how many days you want to stay in your destination(s). Once the first two steps are completed, you need to find out the activities during your trip. Last but not least, you need to decide on your budget.
Once all of these factors are decided, you should work backward and spread your budget between accommodations, activities and means of transportation.
|
|
null | false
| null |
Why do we grow two sets of teeth?
|
It might seem odd to replace a complete set of teeth, but your first set of milk teeth has done its job by the time you are five or six years old. Milk teeth have helped you chew for essential nourishment and learn to talk when it's time for your body to house a more extensive adult set of teeth. You lose milk teeth and get adult teeth.
|
|
null | false
| null |
Virat Kohli (Hindi pronunciation: [ʋɪˈɾɑːʈ ˈkoːɦli] (listen); born 5 November 1988) is an Indian international cricketer and the former captain of the Indian national cricket team who plays as a right-handed batsman for Royal Challengers Bangalore in the IPL and for the Delhi in Indian domestic cricket. Widely regarded as one of the greatest batsmen of all time, Kohli holds the records for scoring most runs in T20 internationals and in the IPL. In 2020, the International Cricket Council named him the male cricketer of the decade. Kohli has also contributed to a number of India's successes, including winning the 2011 World Cup and the 2013 Champions trophy.
Born and raised in New Delhi, Kohli trained at the West Delhi Cricket Academy and started his youth career with the Delhi Under-15 team. He made his international debut in 2008 and quickly became a key player in the ODI team and later made his Test debut in 2011. In 2013, Kohli reached the number one spot in the ICC rankings for ODI batsmen for the first time. During 2014 T20 World Cup, he set a record for the most runs scored in the tournament. In 2018, he achieved yet another milestone, becoming the world's top-ranked Test batsman, making him the only Indian cricketer to hold the number one spot in all three formats of the game. His form continued in 2019, when he became the first player to score 20,000 international runs in a single decade. In 2021, Kohli made the decision to step down as the captain of the Indian national team for T20Is, following the T20 World Cup and in early 2022 he stepped down as the captain of the Test team as well.
|
From the passage mention the years in which Virat Kohli won a trophy. Display the results in comma separated format.
|
2011, 2013
|
null | false
| null |
Give me a list of the Mercury Prize winning albums, since 2000
|
- 'The Hour of Bewilderbeast' by Badly Drawn Boy
- 'Stories from the City, Stories from the Sea' by PJ Harvey
- 'A Little Deeper' by Ms. Dynamite
- 'Boy in da Corner' by Dizzee Rascal
- 'Franz Ferdinand' by Franz Ferdinand
- 'I Am a Bird Now' by Antony and the Johnsons
- 'Whatever People Say I Am, That's What I'm Not' by Arctic Monkeys
- 'Myths of the Near Future' by Klaxons
- 'The Seldom Seen Kid' by Elbow
- 'Speech Therapy' by Speech Debelle
- 'xx' by The xx
- 'Let England Shake' by PJ Harvey
- 'An Awesome Wave' by alt-J
- 'Overgrown' by James Blake
- 'Dead' by Young Fathers
- 'At Least For Now' by Benjamin Clementine
- 'Konnichiwa' by Skepta
- 'Process' by Sampha
- 'Visions of a Life' by Wolf Alice
- 'Psychodrama' by Dave
- 'Kiwanuka' by Michael Kiwanuka
- 'Collapsed in Sunbeams' by Arlo Parks
- 'Sometimes I Might Be Introvert' by Little Simz
|
|
null | false
| null |
Is a blackberry a fruit or a berry?
|
While blackberry sounds like a berry, it's actually a fruit. To be more accurate, it's considered an "aggregate fruit," a fruit that consists of multiple smaller fruits. It's made of multiple small drupelets that are arranged around a central core.
|
|
null | false
| 159
|
The generators were implemented using the TensorFlow library BIBREF31 and trained with training, validation and testing ratio as 3:1:1. The hidden layer size, beam size were set to be 80 and 10, respectively, and the generators were trained with a $70\%$ of dropout rate. We performed 5 runs with different random initialization of the network and the training is terminated by using early stopping. We then chose a model that yields the highest BLEU score on the validation set as shown in Table 2 . Since the trained models can differ depending on the initialization, we also report the results which were averaged over 5 randomly initialized networks. Note that, except the results reported in Table 2 , all the results shown were averaged over 5 randomly initialized networks. We set $\lambda $ to 1000 to severely discourage the reranker from selecting utterances which contain either redundant or missing slots. For each DA, we over-generated 20 candidate sentences and selected the top 5 realizations after reranking. Moreover, in order to better understand the effectiveness of our proposed methods, we: (i) performed an ablation experiments to demonstrate the contribution of each proposed cells (Tables 2 , 3 ), (ii) trained the models on the Laptop domain with varied proportion of training data, starting from $10\%$ to $100\%$ (Figure 3 ), (iii) trained general models by merging all the data from four domains together and tested them in each individual domain (Figure 4 ), and (iv) trained adaptation models on merging data from restaurant and hotel domains, then fine tuned the model on laptop domain with varied amount of adaptation data (Figure 5 ).
We performed 5 runs with different random initialization of the network and the training is terminated by using early stopping.
|
How many runs are performed before stopping?
|
They performed 5 runs with different random initialization of the network.
|
null | false
| null |
During the 78 era, when the maximum capacity of a single ordinary record side or cylinder was less than five minutes, Spalding recorded extensively for Edison Records, with some issues on cylinders and many more on diamond discs. Most featured short works or encore pieces that could fit on a single record side. These recordings were all by the acoustical process, as well as vertically-cut, through 1925, but he made his first electrical recordings in 1926 for Brunswick Records using that company's problematic "Light-Ray" system. After his unsatisfactory experience with Brunswick, Spalding went back to Edison and made some electrical Edison hill-and-dale Diamond Discs as well as a very few Edison "Needle Cut" lateral recordings in late 1928. These were much better recorded than Spalding's Brunswicks, but the Diamond Discs sold as scantily as the rest of Edison's product in that period, and the "Needle-Cut" discs were issued only for a very short time—from August to November 1929—and are exceedingly rare today. Following the Edison company's demise in November 1929, he recorded a handful of more extended works broken across multiple sides for RCA Victor Records.
|
Please give a summary of the type of recordings Spalding made by year.
|
1925: Vertically-cut acoustic recordings
1926: First electrical recordings for Brunswick Records
1928: Electrical recordings using Edison hill-and-dale Diamond Discs and "Needle Cut" recordings.
|
null | false
| null |
Jacque Fresco was born on March 13, 1916, and grew up in a Sephardi Jewish household, at the family's home in Bensonhurst, in the Brooklyn borough of New York City. Fresco's father was an agriculturist born in Constantinople (now Istanbul, Turkey), while his mother Lena was an emigrant from Jerusalem. He later turned his attention to technocracy. A teenager during the Great Depression, he spent time with friends discussing Charles Darwin, Albert Einstein, science, and the future. Fresco attended the Young Communist League before being "physically ejected" for loudly stating that "Karl Marx was wrong!" after a discussion with the league president during a meeting. He left home at the age of 14, hitchhiking and "jumping" trains as one of the so-called "Wild Boys of the Road".
|
Summarize Jacque Fresco's early life based on this passage.
|
Jacque Fresco was born into a Jewish home in Brooklyn in 1916. His teenage years coincided with the Great Depression where he spent his time discussing historically notable thinkers with his friends, such as Albert Einstein and Charles Darwin. He was involved in youth communist groups but eventually left over ideological disputes. He eventually became one of the "Wild Boys of the Road" when 14 and left behind his home.
|
null | false
| null |
Put yourself in the place of a consumer who is considering the purchase of a new car. What sources of information would you use to evaluate your options during ZMOT (Zero moment of truth)?
|
Car brand websites:
After deciding on a car category (compact SUV), I would visit the different car brand websites to understand the offering. I would start comparing the car's various attributes, like engine/battery power, mileage/range, colors, upgrade options, safety features, etc., from the website.
YouTube reviews:
I would watch a few car review video guides on YouTube to speed up the car-buying process and narrow down the list to a few options.
Price estimates:
Once I narrowed it down to a few options, I would obtain a price estimate by sourcing information from KBB.com or TrueCar.com.
Check dealer inventory & quotes:
With a price estimate, I would start contacting the dealers or configure to price (for Tesla). Alternatively, I would keep checking the dealer’s inventory to see if there is an overstock of a particular car color/model for which I can get a better deal.
|
|
null | false
| null |
Why is biathlon a challenging sport?
|
Biathlon requires two very different skills from athletes. First, they need to be fast on skis, hence their heart rate will go high. But second, they also need to be very precise at shooting, hence be able to lower their heart rate very fast. Having both skills and being able to transition from high intensity to very low intensity activities fast is extremely challenging.
|
|
null | false
| null |
In the series, A Song of Ice and Fire, who is the founder of House Yew?
|
Alan o' the Oak, also known as "the blind Bowman"
|
|
null | false
| null |
While pursuing his doctorate in computer science, Metcalfe took a job with MIT's Project MAC after Harvard refused permission for him to connect the university to the then-new ARPAnet. At MAC, Metcalfe was responsible for building some of the hardware that would link MIT's minicomputers with ARPAnet. Metcalfe made ARPAnet the topic of his doctoral thesis, but Harvard initially rejected it. Metcalfe decided how to improve his thesis while working at Xerox PARC, where he read a paper about the ALOHA network at the University of Hawaii. He identified and fixed some of the bugs in the AlohaNet model, then added that work to his revised thesis. It was then accepted by Harvard, which granted his PhD.
Metcalfe was working at PARC in 1973 when he and David Boggs invented Ethernet, initially as a standard for connecting computers over short distances. He later recalled that Ethernet was born on May 22, 1973, the day he circulated a memo titled "Alto Ethernet" which contained a rough schematic of how it would work. "That is the first time Ethernet appears as a word, as does the idea of using coax as ether, where the participating stations, like in AlohaNet or ARPAnet, would inject their packets of data, they'd travel around at megabits per second, there would be collisions, and retransmissions, and back-off," Metcalfe explained. Boggs argued that another date was the birth of Ethernet: November 11, 1973, the first day the system actually functioned.
In 1979, Metcalfe departed PARC and co-founded 3Com, a manufacturer of computer networking equipment, in his Palo Alto apartment. 3Com became a leading provider of networking solutions, and Ethernet became the dominant networking standard for local area networks (LANs). In 1980 he received the ACM Grace Hopper Award for his contributions to the development of local networks, specifically Ethernet. In 1990, the 3Com board of directors appointed Éric Benhamou as CEO instead of Metcalfe, who then left the company. He spent 10 years as a publisher and pundit, writing an internet column for InfoWorld. In 1996, he co-founded Pop!Tech, an executive technology conference. He became a venture capitalist in 2001 and subsequently a general partner at Polaris Venture Partners.
From 2011 to 2021 he was a professor at The University of Texas at Austin's Cockrell School of Engineering, specializing in innovation initiatives. Metcalfe was a keynote speaker at the 2016 Congress of Future Science and Technology Leaders and in 2019 he presented the Bernard Price Memorial Lecture in South Africa. In June 2022, Metcalfe returned to MIT by joining the Computer Science and Artificial Intelligence Laboratory as a research affiliate and computational engineer, working with the MIT Julia Lab.
|
What was Robert Metcalfe's impact on society?
|
Robert Metcalfe was a successful entrepreneur, having started companies like 3Com and Pop!Tech. He was also a professor at the University of Texas and a researcher at the Massachusetts Institute of Technology. His biggest contribution, however, was the invention of Ethernet in 1973. This became the standard for wired internet connection, which is still used 50 years later.
|
null | false
| null |
The Narmada River, also called the Reva and previously also known as Narbada or anglicised as Nerbudda, is the 5th longest river and overall longest west-flowing river in India. It is also the largest flowing river in the state of Madhya Pradesh. This river flows through the states of Madhya Pradesh and Gujarat in India. It is also known as the "Life Line of Madhya Pradesh and Gujarat" due to its huge contribution to the two states in many ways. The Narmada River rises from the Amarkantak Plateau in Anuppur district Madhya Pradesh. It forms the traditional boundary between North India and South India and flows westwards over a length of 1,312 km (815.2 mi) before draining through the Gulf of Khambhat into the Arabian Sea, 30 km (18.6 mi) west of Bharuch city of Gujarat.
|
Tell me some key points about Narmada river from the given text.
|
1.The Narmada is the largest flowing river in madhya pradesh.
2.The Narmada is also known as Rewa or Narbada.
3.It is also known as life line of madhya pradesh and Gujarat.
4.The Narmada rises from the Amarkantak (m.p.)and drain into Arabian sea through the gulf of khambat.
5. It flows westwards over a length of 1312 km & forms boundary between North India & South India.
|
null | false
| null |
Can you help me avoid credit card debt?
|
Here are some strategies to avoid credit card debt:
1. Set a monthly spending budget and stay within it.
2. Always pay off your entire credit card bill each month.
3. Identify areas where you can reduce your spend (eating out, gifts, etc.)
|
|
null | false
| null |
What are the top semiconductor companies?
|
1. Intel
2. AMD
3. NVIDIA
4. Micron
5. Texas Instruments
6. Samsung
7. TSMC
8. Analog Devices
9. National Semiconductors
10. Onsemi
|
|
null | false
| null |
What is the best way to get from London to Paris.
|
To travel from London to Paris you have several options. You can fly from one of London's airports (Heathrow, Gatwick, Stansted or Luton) to Charles de Gaulle or Orly, you could drive crossing the channel from Dover to Calais by ferry or Eurotunnel. The best option is almost certainly to take the Eurostar train from London St Pancras International to Gare du Nord.
|
|
null | false
| 88
|
We use two popular knowledge bases: WordNet BIBREF0 and Freebase BIBREF1 in this paper. Specifically, we use WN18 (a subset of WordNet) BIBREF24 and FB15K (a subset of Freebase) BIBREF2 since their text descriptions are easily publicly available. Table 1 lists statistics of the two datasets.
We use two popular knowledge bases: WordNet [Miller, 1995] and Freebase [Bollacker et al., 2008] in this paper. Specifically, we use WN18 (a subset of WordNet) [Bordes et al., 2014] and FB15K (a subset of Freebase) [Bordes et al., 2013] since their text descriptions are easily publicly avail- able.
|
What dataset is used for the task?
|
WN18 (a subset of WordNet) [Bordes et al., 2014] and FB15K (a subset of Freebase) [Bordes et al., 2013]
|
null | false
| null |
Carbonara (Italian: [karboˈnaːra]) is a Roman pasta dish made with eggs, hard cheese, cured pork and black pepper. The dish took its modern form and name in the middle of the 20th century.
The cheese is usually Pecorino Romano, Parmigiano-Reggiano, or a combination of the two. Spaghetti is the most common pasta, but fettuccine, rigatoni, linguine, or bucatini are also used. Normally guanciale or pancetta are used for the meat component, but lardons of smoked bacon are a common substitute outside Italy.
|
List the types of pasta that can be used to make carbonara
|
Spaghetti, fettuccine, rigatoni, linguine, bucatini
|
null | false
| null |
What is bicycling and why would someone do it? Is bicycling healthy and joyful or is it painful and a waste of time? How many bicycles should you have? Which part of the world do you find the most people bicycling?
|
Bicycling is a form of human powered movement that typically involves sitting or standing on a machine with two wheels that can be propelled forward by using one's arms or legs. The bicycle was invented in the early 1800 (1817) by a German inventor named Drais. The early bicycles were built from wood and soon after used heavy steel for the frame and wheels. These steel bikes were called Boneshakers because when you road on them you would feel every stone and bump through your bones. Bicycling exists in nearly every corner of the world, on every continent and it is done by all people. In many countries, bicycling is the primary mode of transportation for people, especially in urban settings. Bicycling is also a professional sporting competition where men and women race against one another in large groups in front of cheering crowds. Bicycling is a very efficient form of converting power generated by a human into forward motion. While most humans can't run faster than 10 miles per hour for more than a few miles, humans can go much faster and often hundreds of miles when using a bicycle. Because modern bicycles use pneumatic tires and have many gears, it is considered a very healthy form of exercise. People who love bicycling often have many different bicycles for different conditions, such as dirt or gravel, pavement, snow and even sand.
|
|
null | false
| 319
|
Visual Question Answering (VQA) refers to a challenging task which lies at the intersection of image understanding and language processing. The VQA task has witnessed a significant progress in the recent years by the machine intelligence community. The aim of VQA is to develop a system to answer specific questions about an input image. The answer could be in any of the following forms: a word, a phrase, binary answer, multiple choice answer, or a fill in the blank answer. Agarwal et al. BIBREF0 presented a novel way of combining computer vision and natural language processing concepts of to achieve Visual Grounded Dialogue, a system mimicking the human understanding of the environment with the use of visual observation and language understanding.
The advancements in the field of deep learning have certainly helped to develop systems for the task of Image Question Answering. Krizhevsky et al BIBREF1 proposed the AlexNet model, which created a revolution in the computer vision domain. The paper introduced the concept of Convolution Neural Networks (CNN) to the mainstream computer vision application. Later many authors have worked on CNN, which has resulted in robust, deep learning models like VGGNet BIBREF2, Inception BIBREF3, ResNet BIBREF4, and etc. Similarly, the recent advancements in natural language processing area based on deep learning have improved the text understanding prforance as well. The first major algorithm in the context of text processing is considered to be the Recurrent Neural Networks (RNN) BIBREF5 which introduced the concept of prior context for time series based data. This architecture helped the growth of machine text understanding which gave new boundaries to machine translation, text classification and contextual understanding. Another major breakthrough in the domain was the introduction of Long-Short Term Memory (LSTM) architecture BIBREF6 which improvised over the RNN by introducing a context cell which stores the prior relevant information.
The vanilla VQA model BIBREF0 used a combination of VGGNet BIBREF2 and LSTM BIBREF6. This model has been revised over the years, employing newer architectures and mathematical formulations. Along with this, many authors have worked on producing datasets for eliminating bias, strengthening the performance of the model by robust question-answer pairs which try to cover the various types of questions, testing the visual and language understanding of the system. In this survey, first we cover major datasets published for validating the Visual Question Answering task, such as VQA dataset BIBREF0, DAQUAR BIBREF7, Visual7W BIBREF8 and most recent datasets up to 2019 include Tally-QA BIBREF9 and KVQA BIBREF10. Next, we discuss the state-of-the-art architectures designed for the task of Visual Question Answering such as Vanilla VQA BIBREF0, Stacked Attention Networks BIBREF11 and Pythia v1.0 BIBREF12. Next we present some of our computed results over the three architectures: vanilla VQA model BIBREF0, Stacked Attention Network (SAN) BIBREF11 and Teney et al. model BIBREF13. Finally, we discuss the observations and future directions.
Visual Question Answering (VQA) refers to a challenging task which lies at the intersection of image understanding and language processing. The VQA task has witnessed a significant progress the recent years by the machine intelligence community. The aim of VQA is to develop a system to answer specific questions about an input image.
|
What is the aim of VQA?
|
To develop a system to answer specific questions about an input image.
|
null | false
| null |
Can I make chocolate at home?
|
Making chocolate at home can be rewarding, but be prepared to spend a little money and a lot of time. The money goes to equipment and cocoa beans. For equipment, you will need a melanger (about $600) to grind the beans and a tempering machine (about $400) to make the finished product. You will also need a silicone mat (about $20) and a hair dryer (about $30). But let's start at the beginning.
Buy between two and six pounds of fermented raw cocoa beans. There are a variety of retail importers you can find on the web. Choose a single variety or a blend that appeals to your taste in chocolate.
Sort the beans, removing broken beans, badly misshapen beans, and whatever foreign bodies may be present.
Roast the beans, two pounds at a time, at 350F for between 20 and 30 minutes, depending on your experience with the beans or by how they smell. They will smell like fresh brownies. Don't scorch them!
Let the beans cool.
Peel the beans. You can do this by hand - which I find produces a superior product - but be warned that this can take up to 6 hours per pound. Alternatively, you can purchase a winnower and use the hair dryer to blow out the skins. This approach is much faster, but leaves more skin and loses more bean fragments. Either way, weigh your final bean yield.
Get you melanger going. Use the hair dryer to get the grinding surfaces up to at least 120F. Add the beans a handful at a time. Help the melanger along if it gets stuck. Make sure the melanger is going steadily before leaving it unattended. When it gets going, it will flow well.
Run the melanger for 24 to 72 hours, again depending on your experience with the beans and your tastes. If trying a mix of beans, add the more robustly flavored beans early in the process and the more delicately flavored beans later. Give even the most delicate beans at least 8 hours. Add sugar to taste. The percentage of sugar by weight can vary from a few percent up to 30 percent. You may also add dried whole fat milk at this point, and other flavorings, like an inch or so of vanilla bean. Melange the sugar and other additives at least 8 more hours. The goal is to decrease the particle size below what can be detected by the tongue. When done, decant into a silicone mat line baking pan. This is your untempered chocolate. it melts easily and tends to get crumbly as it ages. Which is why you will want to temper it.
Finally, follow the directions on your tempering machine, adding chunks of untempered chocolate as indicated. The liquid tempered chocolate can be used for fruit dipping or can be poured into molds for bars or shaped candies. Other additives can be included at this stage - nuts, nougat etc. Cool the results at room temperature for several hours before serving. Enjoy the results of your labor.
|
|
null | false
| null |
Name famous French singers (all time).
|
Johnny Hallyday, Michel Sardou, Orelsan, Kendji, Lara Fabian, Jennifer, Charles Aznavour, Edith Piaf.
|
|
null | false
| null |
Why are visualisations useful for communicating information?
|
There are several reasons that visualisations are useful for communicating information. One reason is that the messages we are trying to convey can often be based on large amounts of data. In this case, visualisations can provide a useful way of summarising the data to present significant patterns in an easily consumable manner. Another reason worth highlighting is that the choice of the right visualisations can direct the viewer to a conclusion, for example a heat-map might concentrate the viewers attention to a particular set of data points.
|
|
null | false
| null |
As established by law and government policy, the two official forms of written Norwegian are Bokmål (literally 'book tongue') and Nynorsk ('new Norwegian'). The official Norwegian Language Council (Språkrådet) is responsible for regulating the two forms, and recommends the terms Norwegian Bokmål and Norwegian Nynorsk in English.[citation needed] Two other written forms without official status also exist. One, called Riksmål ('national language'), is today to a large extent the same language as Bokmål though somewhat closer to the Danish language. It is regulated by the unofficial Norwegian Academy, which translates the name as 'Standard Norwegian'. The other is Høgnorsk ('High Norwegian'), a more purist form of Nynorsk, which maintains the language in an original form as given by Ivar Aasen and rejects most of the reforms from the 20th century; this form has limited use.
|
What are the unofficial forms of written Norwegian ?
|
- Riksmål
- Høgnorsk
|
null | false
| null |
What are examples of endocrine disruptors?
|
BPA (bisphenol-A), phthalates, parabens and xenoestrogens are some examples of EDC's (endocrine disrupting chemicals).
|
|
null | false
| null |
What are recommended things to do or have when adopting a puppy?
|
Adopting a puppy is a lot of fun, but can also be a lot of work. The bare essentials will be necessary: food, water, and a living space for your puppy. Many dog owners opt to have a crate for their puppy which can be a place where the puppy can sleep and be alone, away from outside stimuli. Furthermore, many pet owners will also get pet insurance which can help cover any unforeseen medical issues or conditions outside of the normal vaccinations a puppy typically requires. Recreational items are also highly encouraged. Things like toys, treats, and collar with leash are highly recommended. If you are aware of the upfront costs and requirements for raising a puppy, the experience will be a lot better for owners and the dog!
|
|
null | false
| 356
|
In this section, we provide comparative results on our new Multilingual Document Classification Corpus. Since the initial work by BIBREF0 many alternative approaches to cross-lingual document classification have been developed. We will encourage the respective authors to evaluate their systems on MLDoc. We believe that a large variety of transfer language pairs will give valuable insights on the performance of the various approaches.
In this paper, we propose initial strong baselines which represent two complementary directions of research: one based on the aggregation of multilingual word embeddings, and another one, which directly learns multilingual sentence representations. Details on each approach are given in section "Multilingual word representations" and "Multilingual sentence representations" respectively. In contrast to previous works on cross-lingual document classification with RVC2, we explore training the classifier on all languages and transfer it to all others, ie. we do not limit our study to the transfer between English and a foreign language.
One can envision several ways to define cross-lingual document classification, in function of the resources which are used in the source and transfer language (see Table 3 ). The first scheme assumes that we have no resources in the transfer language at all, neither labeled nor unlabeled. We will name this case “zero-shot cross-lingual document classification”. To simplify the presentation, we will assume that we transfer from English to German. The training and evaluation protocol is as follows. First, train a classifier using resources in the source language only, eg. the training and development corpus are in English. All meta parameters and model choices are performed using the English development corpus. Once the best performing model is selected, it is applied to the transfer language, eg. the German test set. Since no resources of the transfer language are used, the same system can be applied to many different transfer languages. This type of cross-lingual document classification needs a very strong multilingual representation since no knowledge on the target language was used during the development of the classifier.
In a second class of cross-lingual document classification, we may aim in improving the transfer performance by using a limited amount of resources in the target language. In the framework of the proposed MLDoc we will use the development corpus of target language for model selection. We will name this method “targeted cross-lingual document classification” since the system is tailored to one particular transfer language. It is unlikely that this system will perform well on other languages than the ones used for training or model selection.
If the goal is to build one document classification system for many languages, it may be interesting to use already several languages during training and model selection. To allow a fair comparison, we will assume that these multilingual resources have the same size than the ones used for zero-shot or targeted cross-language document classification, e.g. a training set composed of five languages with 200 examples each. This type of training is not a cross-lingual approach any more. Consequently, we will refer to this method as “joint multilingual document classification”.
In this paper, we propose initial strong baselines which represent two complementary directions of research: one based on the aggregation of multilingual word embeddings, and another one, which directly learns multilingual sentence representations.
|
What are the initial strong baselines which represent two complementary directions of research?
|
One based on the aggregation of multilingual word embeddings, and another one, which directly learns multilingual sentence representations.
|
null | false
| 267
|
In this work we present an ensemble classifier that is detecting hate-speech in short text, such as tweets. The input to the base-classifiers consists of not only the standard word uni-grams, but also a set of features describing each user's historical tendency to post abusive messages. Our main innovations are: i) a deep learning architecture that uses word frequency vectorisation for implementing the above features, ii) an experimental evaluation of the above model on a public dataset of labeled tweets, iii) an open-sourced implementation built on top of Keras.
The results show that our approach outperforms the current state of the art, and to the best of our knowledge, no other model has achieved better performance in classifying short messages. The approach does not rely on pre-trained vectors, which provides a serious advantage when dealing with short messages of this kind. More specifically, users will often prefer to obfuscate their offensive terms using shorter slang words or create new words by `inventive' spelling and word concatenation. For instance, the word `Islamolunatic' is not available in the popular pre-trained word embeddings (Word2Vec or GloVe), even though it appears with a rather high frequency in racist postings. Hence, word frequency vectorization is preferable to the pre-trained word embeddings used in prior works if one aims to build a language-agnostic solution.
We believe that deep learning models have a high potential wrt. classifying text or analyzing the sentiment in general. In our opinion there is still space for further improving the classification algorithms.
For future work we plan to investigate other sources of information that can be utilized to detect hateful messages. In addition, we intend to generalize the output received in the current experiment, with evaluation over other datasets, including analyzing texts written in different languages.
For future work we plan to investigate other sources of information that can be utilized to detect hateful messages. In addition, we intend to generalize the output received in the current experiment, with evaluation over other datasets, including analyzing texts written in different languages.
|
Do they have any plans for future studies?
|
They plan to investigate other sources of information that can be utilized to detect hateful messages. In addition, they intend to generalize the output received in the current experiment, with evaluation over other datasets, including analyzing texts written in different languages.
|
null | false
| null |
The Ford Telstar is an automobile that was sold by Ford in Asia, Australasia and Africa, comparable in size to the European Ford Sierra and the North American Ford Tempo. It was progressively replaced by the Ford Mondeo. It was named after the Telstar satellite.
|
Based on this paragraph, where was the The Ford Telstar sold?
|
Asia, Australasia and Africa
|
null | false
| null |
Penguin Books is a British publishing house. It was co-founded in 1935 by Allen Lane with his brothers Richard and John, as a line of the publishers The Bodley Head, only becoming a separate company the following year. Penguin revolutionised publishing in the 1930s through its inexpensive paperbacks, sold through Woolworths and other stores for sixpence, bringing high-quality fiction and non-fiction to the mass market. Its success showed that large audiences existed for serious books. It also affected modern British popular culture significantly through its books concerning politics, the arts, and science.
|
What is Penguin Books?
|
Penguin Books is a publishing house of inexpensive fiction and non-fiction paperbacks. It was founded by the Lane brothers in 1935 as a line of the publishers The Bodley Head. It became a separate publisher in 1936.
|
null | false
| null |
Artificial intelligence (AI) is intelligence demonstrated by machines, as opposed to intelligence of humans and other animals. Example tasks in which this is done include speech recognition, computer vision, translation between (natural) languages, as well as other mappings of inputs.
AI applications include advanced web search engines (e.g., Google Search), recommendation systems (used by YouTube, Amazon, and Netflix), understanding human speech (such as Siri and Alexa), self-driving cars (e.g., Waymo), generative or creative tools (ChatGPT and AI art), automated decision-making, and competing at the highest level in strategic game systems (such as chess and Go).
As machines become increasingly capable, tasks considered to require "intelligence" are often removed from the definition of AI, a phenomenon known as the AI effect. For instance, optical character recognition is frequently excluded from things considered to be AI, having become a routine technology.
Artificial intelligence was founded as an academic discipline in 1956, and in the years since it has experienced several waves of optimism, followed by disappointment and the loss of funding (known as an "AI winter"), followed by new approaches, success, and renewed funding. AI research has tried and discarded many different approaches, including simulating the brain, modeling human problem solving, formal logic, large databases of knowledge, and imitating animal behavior. In the first decades of the 21st century, highly mathematical and statistical machine learning has dominated the field, and this technique has proved highly successful, helping to solve many challenging problems throughout industry and academia.
The various sub-fields of AI research are centered around particular goals and the use of particular tools. The traditional goals of AI research include reasoning, knowledge representation, planning, learning, natural language processing, perception, and the ability to move and manipulate objects. General intelligence (the ability to solve an arbitrary problem) is among the field's long-term goals. To solve these problems, AI researchers have adapted and integrated a wide range of problem-solving techniques, including search and mathematical optimization, formal logic, artificial neural networks, and methods based on statistics, probability, and economics. AI also draws upon computer science, psychology, linguistics, philosophy, and many other fields.
The field was founded on the assumption that human intelligence "can be so precisely described that a machine can be made to simulate it". This raised philosophical arguments about the mind and the ethical consequences of creating artificial beings endowed with human-like intelligence; these issues have previously been explored by myth, fiction, and philosophy since antiquity. Computer scientists and philosophers have since suggested that AI may become an existential risk to humanity if its rational capacities are not steered towards beneficial goals. The term artificial intelligence has also been criticized for overhyping AI's true technological capabilities.
|
Please give me summary of the below graph in not more than 2 line.
|
Artificial intelligence (AI) is intelligence demonstrated by machines rather than Intelligence of species and this AI is being used in some application like advanced web search engines, recommendation systems, understanding human speech, self-driving car, generative or creative tools but like every coin has two face AI also raise ethical consequences of creating artificial beings endowed with human-like intelligence
|
null | false
| null |
After listening to a talk by biologist David Baltimore and further stimulation from his virologist colleague, Robert Ting, concerning the work of the late Howard Martin Temin, Gallo became interested in the study of retroviruses, and made their study the primary activity of his lab. In 1976, Doris Morgan, a first year post-doctoral fellow in Gallo's lab, was asked by Gallo to examine culture fluid of activated lymphocytes for the possible production of growth factors. Soon she was successful in growing T lymphocytes. Gallo, Morgan and Frank Ruscetti, another researcher in Gallo's lab, coauthored a paper in Science describing their method. The Gallo group identified this as T-cell growth factor (TCGF). The name was changed in 1978 to IL-2 (interleukin-2) by the Second International Lymphokine Conference (which was held in Interlaken, Switzerland). Although earlier reports had described soluble molecules with biologic effects, the effects and biochemistry of the factors were not well characterized. One such example was the report by Julius Gordon in 1965, which described blastogenic transformation of lymphocytes in extracellular media. However, cell growth was not demonstrated and the affected cell type was not identified, making the identity of the factor(s) involved unclear and its natural function unknown.
The discovery of IL-2 allowed T cells, previously thought to be dead end cells, to be grown significantly in culture for the first time, opening research into many aspects of T cell immunology. Gallo's lab later purified and biochemically characterized IL-2. This breakthrough also allowed researchers to grow T-cells and study the viruses that affect them, such as human T-cell leukemia virus, or HTLV, the first retrovirus identified in humans, which Bernard Poiesz, another post-doctoral fellow in Gallo's lab played a key role in its isolation. HTLV's role in leukemia was clarified when Kiyoshi Takatsuki and other Japanese researchers, puzzling over an outbreak of a rare form of leukemia, later independently found the same retrovirus, and both groups showed HTLV to be the cause. At the same time, a similar HTLV-associated leukemia was identified by the Gallo group in the Caribbean. In 1982, Gallo received the Lasker Award: "For his pioneering studies that led to the discovery of the first human RNA tumor virus [the old name for retroviruses] and its association with certain leukemias and lymphomas."
|
Please give me a short bulleted list of the key discoveries from Gallo’s lab.
|
Successfully growing T lymphocytes in culture for the first time
Interleukin-2 (IL-2)
Human T-cell leukemia virus
The first human retrovirus
|
null | false
| null |
It was initially scripted by screenwriter Carl Foreman, who was later replaced by Michael Wilson. Both writers had to work in secret, as they were on the Hollywood blacklist and had fled to the UK in order to continue working. As a result, Boulle, who did not speak English, was credited and received the Academy Award for Best Adapted Screenplay; many years later, Foreman and Wilson posthumously received the Academy Award.
|
Given this paragraph about the movie 'The Bridge on the River Kwai', why did the writers have to work in secret?
|
They had to work in secret because they were on the Hollywood Blacklist for presumed involvement in the communist movement.
|
1911.03705
| false
| null |
Towards empowering machines with the generative commonsense reasoning ability, we create a large-scale dataset, named CommonGen, for the constrained text generation task. We collect $37,263$ concept-sets as the inputs, each of which contains three to five common concepts. These concept-sets are sampled from several large corpora of image/video captions, such that the concepts inside them are more likely to co-occur in natural scenes. Through crowd-sourcing via Amazon Mechanical Turk (AMT), we finally obtain $89,028$ human-written sentences as expected outputs. We investigate the performance of sophisticated sequence generation methods for the proposed task with both automatic metrics and human evaluation. The experiments show that all methods are far from human performance in generative commonsense reasoning. Our main contributions are as follows: 1) We introduce the first large-scale constrained text generation dataset targeting at generative commonsense reasoning; 2) We systematically compare methods for this (lexically) constrained text generation with extensive experiments and evaluation. 3) Our code and data are publicly available (w/ the URL in the abstract), so future research in this direction can be directly developed in a unified framework.
Following the general definition in the largest commonsense knowledge graph, ConceptNet BIBREF11, we understand a concept as a common noun or verb. We aim to test the ability of generating natural scenes with a given set of concepts. The expected concept-sets in our task are supposed to be likely co-occur in natural, daily-life scenes . The concepts in images/videos captions, which usually describe scenes in our daily life, thus possess the desired property. We therefore collect a large amount of caption sentences from a variety of datasets, including VATEX BIBREF4, LSMDC BIBREF12, ActivityNet BIBREF13, and SNLI BIBREF15, forming 1,040,330 sentences in total.
We collect $37,263$ concept-sets as the inputs, each of which contains three to five common concepts. These concept-sets are sampled from several large corpora of image/video captions, such that the concepts inside them are more likely to co-occur in natural scenes.
The expected concept-sets in our task are supposed to be likely co-occur in natural, daily-life scenes . The concepts in images/videos captions, which usually describe scenes in our daily life, thus possess the desired property. We therefore collect a large amount of caption sentences from a variety of datasets, including VATEX BIBREF4, LSMDC BIBREF12, ActivityNet BIBREF13, and SNLI BIBREF15, forming 1,040,330 sentences in total.
|
Where do the concept sets come from?
|
The answers are shown as follows:
* These concept-sets are sampled from several large corpora of image/video captions
|
null | false
| null |
San Francisco International Airport (IATA: SFO, ICAO: KSFO, FAA LID: SFO) is an international airport in an unincorporated area of San Mateo County, 13 miles (21 km) south of Downtown San Francisco. It has flights to points throughout North America and is a major gateway to Europe, the Middle East, Asia, and Oceania.
SFO is the largest airport in the San Francisco Bay Area and the second-busiest in California, after Los Angeles International Airport (LAX). In 2017, it was the seventh-busiest airport in the United States and the 24th-busiest in the world by passenger count. It is the fifth-largest hub for United Airlines, which operates out of Terminal 3 and the International Terminal. SFO functions as United's primary transpacific gateway. Additionally, SFO is a major maintenance hub for United Airlines and houses the SFO Museum that was created in 1980, the first museum in an international airport. It also serves as a hub for Alaska Airlines, which operates in Terminal 2.
The airport is owned and operated by the City and County of San Francisco and lies in a ZIP code assigned to San Francisco. However, it is located outside of San Francisco in unincorporated San Mateo County. Between 1999 and 2004, the San Francisco Airport Commission operated city-owned SFO Enterprises, Inc., to oversee its business purchases and operations of ventures.
|
What are the statistics about San Francisco International Airport?
|
It is an international airport 13 miles south of San Francisco. It is the largest airport in San Francisco Bay Area and second largest in the state of California. It is a 7th business airport in United States and 24th in the world by passenger count. It is one of the largest hub for United Airlines. it houses the SFO Museum which was created in 1980. It is owned and operated by the city and county of San Francisco
|
1911.00473
| false
| null |
To fine-tune BERT, we used a proprietary corpus that consists of hundreds of thousands of legal agreements. We extracted text from the agreements, tokenized it into sentences, and removed sentences without alphanumeric text. We selected the BERT-Base uncased pre-trained model for fine-tuning. To avoid including repetitive content found at the beginning of each agreement we selected the 31st to 50th sentence of each agreement. We ran unsupervised fine-tuning of BERT using sequence lengths of 128, 256 and 512. The loss function over epochs is shown in Figure FIGREF3.
To fine-tune BERT, we used a proprietary corpus that consists of hundreds of thousands of legal agreements.
|
How big is dataset used for fine-tuning BERT?
|
The answers are shown as follows:
* hundreds of thousands of legal agreements
|
null | false
| 401
|
Medical providers across the United States are required to document clinical visits in the Electronic Health Records. This need for documentation takes up a disproportionate amount of their time and attention, resulting in provider burnout BIBREF0, BIBREF1. One study found that full-time primary care physicians spent about 4.5 hours of an 11-hour workday interacting with the clinical documentation systems, yet were still unable to finish their documentation and had to spend an additional 1.4 hours after normal clinical hours BIBREF2.
Speech and natural language processing are now sufficiently mature that there is considerable interest, both in academia and industry, to investigate how these technologies can be exploited to simplify the task of documentation, and to allow providers to dedicate more time to patients. While domain-specific automatic speech recognition (ASR) systems that allow providers to dictate notes have been around for a while, recent work has begun to address the challenges associated with generating clinical notes directly from speech recordings. This includes inducing topic structure from conversation data, extracting relevant information, and clinical summary generation BIBREF3. In one recent work, authors outlined an end-to-end system; however, the details were scant without empirical evaluations of their building blocks BIBREF4. One of the simplistic approaches uses a hand crafted finite state machine based grammar to locate clinical entities in the ASR transcripts and map them to canonical clinical terms BIBREF5. This seems to perform well in a narrowly scoped task. A more ambitious approach mapped ASR transcripts to clinical notes by adopting a machine translation approach BIBREF6. However this performed poorly. To address the difficulty in accessing clinical data, researchers have experimented with synthetic data to develop a system for documenting nurse-initiated telephone conversations for congestive heart failure patients who are undergoing telemonitoring after they have been discharged from the hospital BIBREF7. In their task, a question-answer based model achieved an F-score of 0.80. This naturally raises the question of how well state-of-art techniques will perform in helping the broader population of clinicians such as primary care providers.
One might expect that the task of extracting clinical concepts from audio faces challenges similar to the domain of unstructured clinical texts. In that domain, one of the earliest public-domain tasks is the i2b2 relations challenge, defined on a small corpus of written discharge summaries consisting of 394 reports for training, 477 for test, and 877 for evaluation BIBREF8. Given the small amount of training data, not surprisingly, a disproportionately large number of teams fielded rule-based systems. Conditional random field-based (CRF) systems BIBREF9 however did better even with the limited amount of training data BIBREF10. Other i2b2/n2c2 challenges focused on coreference resolution BIBREF11, temporal relation extraction BIBREF12, drug event extraction BIBREF13 on medical records, and extracting family history BIBREF14. Even though the text was largely unstructured, they benefited from punctuation and capitalization, section headings and other cues in written domain which are unavailable in audio to the same extent.
With the goal of creating an automated medical scribe, we broke down the task into modular components, including ASR and speaker diarization which are described elsewhere BIBREF15. In this work, we investigate the task of extracting relevant clinical concepts from transcripts. Our key contributions include: (i) defining three tasks – the Medications Task, the Symptoms Task, and the Conditions Task along with principles employed in developing the annotation guidelines for them (Section SECREF2); (ii) measuring the label quality using inter-labeler agreements and refining the quality iteratively (Section SECREF3), (iii) evaluating the performance of the state-of-the-art models on these tasks (Section SECREF4), and (iv) a comprehensive analysis of the performance of the models including manual error categorization (Section SECREF5). The corpus we have created in this work is based on private, proprietary data that cannot be publicly shared. Instead, we are sharing the learnings from our experience that might be useful for the wider community as well as the detailed labeling guidelines as supplementary material in the extended version of this paper on arxiv.org.
In this work, we investigate the task of extracting relevant clinical concepts from transcripts.
|
What task do they investigate in the paper?
|
The task of extracting relevant clinical concepts from transcripts.
|
null | false
| null |
The spelling expresso is mostly considered incorrect, though some sources call it a less common variant. It is common in French and Portuguese. Italy uses the term espresso, substituting s for most x letters in Latin-root words; x is not considered part of the standard Italian alphabet. Italian people commonly refer to it simply as caffè (coffee), espresso being the ordinary coffee to order; the same happens in Portugal (café), with some regional variations (bica in Lisbon and cimbalino traditionally in Porto); in Spain, while café expreso is seen as the more "formal" denomination, café solo (alone, without milk) is the usual way to ask for it when at an espresso bar.
|
Using valid, well-formatted JSON, generate a list of espresso spellings based on the text provided
|
[
"espresso",
"expresso",
"coffee",
"caffè",
"café",
"bica",
"cimbalino",
"café expreso",
"café solo"
]
|
null | false
| null |
Sogndal is a municipality in Vestland county, Norway. It is located on the northern shore of the Sognefjorden in the traditional district of Sogn. The village of Hermansverk is the administrative center of Sogndal municipality. Other villages include Kaupanger, Kjørnes, Fimreite, Nornes, and Fjærland. Sogndal Airport, Haukåsen is located 10 kilometres (6.2 mi) southwest of Kaupanger.
|
What is Sogndal?
|
Sogndal is a small city in southwestern Norway with a population of about 12,000
|
null | false
| null |
Are these things typically yellow, green or red: the sun, a banana, blood, lettuce, a tomato, grass, peas, and a heart emoji.
|
The items that are usually yellow are the sun and a banana. The red items are blood, a tomato, and a heart emoji. The green items are lettuce, grass and peas.
|
|
null | false
| null |
Barthélemy Hus, called Hus-Desforges (18 July 1699 in Bordeaux – 1 September 1786 in Lyon, aged 63) was an 18th-century French comedian and troupe leader.
The son of maître à danser Jérôme Hus and Marguerite Pageot, called Desforges, he was the youngest member of the Hus family and with his brother François, directed a touring company known as the "Hus brothers troupe."
Around 1722, he married Marie Anne Daguerre Ascorette, actress born in Namur (1709-1736). The couple had at least seven children who would be part of the troupe.
On 27 January 1750, he married in Brussels his second wife, Louise Vivier de Courtenay, descendant of the famous Courtenay family.
Barthélemy Hus-Desforges was the grandfather of musician Pierre-Louis Hus-Desforges (1773-1838).
|
What was Barthelemy Hus-Desforges famous for
|
Barthelemy Hus, called Hus-Desforges (18 July 1699 in Bordeaux – 1 September 1786 in Lyon, aged 63) was an 18th-century French comedian and troupe leader.
The son of maitre a danser Jerome Hus and Marguerite Pageot, called Desforges, he was the youngest member of the Hus family and with his brother Francois, directed a touring company known as the "Hus brothers troupe."
Around 1722, he married Marie Anne Daguerre Ascorette, actress born in Namur (1709-1736). The couple had at least seven children who would be part of the troupe.
On 27 January 1750, he married in Brussels his second wife, Louise Vivier de Courtenay, descendant of the famous Courtenay family.
Barthelemy Hus-Desforges was the grandfather of musician Pierre-Louis Hus-Desforges (1773-1838).
|
null | false
| null |
Tell me what are these visual charts appropriate for?
Choropleth map, heatmap, scatter plot
Density plot, horizon chart, line chart, Geographic heatmap , Bar chart
|
Although any of these charts might be used for various purposes, here is a list of use cases and corresponding chart types;
Comparison: Bar chart, line chart
Correlation: heatmap, scatter plot
Distribution: Density plot, horizon chart
Geospatial: Choropleth map, Geographic heatmap
|
|
null | false
| null |
Identify all of the breeds of dogs in this list:
Banana
Poodle
Pekingese
Hunan
Apple
Peach
Lettuce
Arubian Cucunu
German Shepherd
Jack Russell Terrier
Wolfhound
Pariah Dog
Tamales
Coquito
Mezcal
Gingko
Papaya
Sequoia
Redwood
Maple
Dogwood
Afghan Hound
Airedale Terrier
Camphor Tree
Eucalyptus
Chihuaha
Basenji
Bichon Frise
Blue Spruce
Magnolia
Cacao
Black Locust
Tequila
|
The following are dog breeds that are found in the list:
Poodle
Pekingese
Arubian Cucunu
German Shepherd
Jack Russell Terrier
Wolfhound
Afghan Hound
Airedale Terrier
Chihuahua
Basenji
Bichon Frise
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.