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
|
|---|---|---|---|---|---|
1911.03894
| false
| null |
We use the unshuffled version of the French OSCAR corpus, which amounts to 138GB of uncompressed text and 32.7B SentencePiece tokens.
We use the unshuffled version of the French OSCAR corpus, which amounts to 138GB of uncompressed text and 32.7B SentencePiece tokens.
|
What is CamemBERT trained on?
|
The answers are shown as follows:
* unshuffled version of the French OSCAR corpus
|
null | false
| null |
Identify the bird from the list: Elder, Eider, Either
|
Eider
|
|
null | false
| null |
Name different types of paper you would see in a notebook
|
Blank
Dotted
Ruled
Squared
|
|
null | false
| null |
Paul McCartney and John Lennon wrote which Rolling Stones song?
|
"I Wanna Be Your Man"
|
|
null | false
| null |
What is a debit card?
|
A debit card, also known as a check card or bank card is a payment card that can be used in place of cash to make purchases. The term plastic card includes the above and as an identity document. These are similar to a credit card, but unlike a credit card, the money for the purchase must be in the cardholder's bank account at the time of a purchase and is immediately transferred directly from that account to the merchant's account to pay for the purchase.
Some debit cards carry a stored value with which a payment is made (prepaid card), but most relay a message to the cardholder's bank to withdraw funds from the cardholder's designated bank account. In some cases, the payment card number is assigned exclusively for use on the Internet and there is no physical card. This is referred to as a virtual card.
|
|
null | false
| null |
disambiguation).
Ramadan
رَمَضَان
Ramadan montage.jpg
From top, left to right: A crescent moon over Sarıçam, Turkey, marking the beginning of the Islamic month of Ramadan. Ramadan Quran reading in Bandar Torkaman, Iran. Community Iftar meal in Dubai, United Arab Emirates, Tarawah prayers in a mosque in West Sumatra, Indonesia. Foods served at a traditional Iftar meal. Ramadan decorations in Jerusalem. Zakat donation boxes in Taipei, Taiwan.
Also called
Albanian: Ramazani
Azerbaijani: Ramazan
Bosnian: Ramazan
Bengali: রমজান, romanized: Rômzan / Rômjan
Hindi: रमज़ान, romanized: Ramzān
Kurdish: ڕەمەزان, romanized: Remezan
Persian: رمضان, romanized: Ramazān
Malay: Puasa
Pashto: روژه, romanized: Rozha
Punjabi: رمضان / ਰਮਜ਼ਾਨ, romanized: Ramzān
Somali: Rabadaan or Rabmadaan
Thai: เราะมะฎอน, romanized: Romadon
Turkish: Ramazan
Sindhi: روزو
Urdu: رمضان, romanized: Ramzān
Zazaki: Remezan
Observed by Muslims
Type Religious
Celebrations Community iftars and Community prayers
Observances
Sawm (fasting)
Zakat and sadaqah (alms giving)
Commemorating Qadr Night
Reading the Quran
Abstaining from all bad deeds and staying humble
Taraweeh prayer (Sunni Muslims)
Begins At the last night of the month of Sha'ban
Ends At the last night of the month of Ramadan
Date Variable (follows the Islamic lunar calendar)
2023 date 22 March – 21 April
Frequency Annual (lunar calendar)
Related to Eid al-Fitr, Qadr Night
Ramadan (Arabic: رَمَضَان, romanized: Ramaḍān [ra.ma.dˤaːn]; also spelled Ramazan, Ramzan, Ramadhan or Ramathan) is the ninth month of the Islamic calendar, observed by Muslims worldwide as a month of fasting (sawm), prayer, reflection and community. A commemoration of Muhammad's first revelation, the annual observance of Ramadan is regarded as one of the Five Pillars of Islam and lasts twenty-nine to thirty days, from one sighting of the crescent moon to the next.
Fasting from dawn to sunset is fard (obligatory) for all adult Muslims who are not acutely or chronically ill, travelling, elderly, breastfeeding, diabetic, or menstruating. The predawn meal is referred to as suhur, and the nightly feast that breaks the fast is called iftar. Although fatwas have been issued declaring that Muslims who live in regions with a midnight sun or polar night should follow the timetable of Mecca, it is common practice to follow the timetable of the closest country in which night can be distinguished from day.
|
What is Ramadan?
|
Ramadan is the holy month of fasting observed by Muslims yearly all over the world.During this month Muslims abstain from food from dawn till dusk for 29 days. Ramadan is also one of the key pillars of Islam.
|
null | false
| null |
In physics and chemistry, the law of conservation of energy states that the total energy of an isolated system remains constant; it is said to be conserved over time. This law, first proposed and tested by Émilie du Châtelet, means that energy can neither be created nor destroyed; rather, it can only be transformed or transferred from one form to another. For instance, chemical energy is converted to kinetic energy when a stick of dynamite explodes. If one adds up all forms of energy that were released in the explosion, such as the kinetic energy and potential energy of the pieces, as well as heat and sound, one will get the exact decrease of chemical energy in the combustion of the dynamite.
|
From the passage provided, extract the name of the person who first proposed and tested the law of conservation of energy.
|
Émilie du Châtelet first proposed and tested the law of conservation of energy.
|
null | false
| 154
|
We present a comprehensive comparison study of the existing corpora for selection-based question answering. Our intrinsic analysis provides a better understanding of the uniqueness or similarity between these corpora. Our extrinsic analysis shows the strength or weakness of combining these corpora together for statistical learning. Additionally, we create a silver-standard dataset for answer retrieval and triggering, which will be publicly available. In the future, we will explore different ways of improving the quality of our silver-standard datasets by fine-tuning the hyper-parameters.
In the future, we will explore different ways of improving the quality of our silver-standard datasets by fine-tuning the hyper-parameters.
|
How to improve the quality of the silver-standard datasets in the future?
|
The authors will explore different ways of improving the quality of the silver-standard datasets by fine-tuning the hyper-parameters.
|
null | false
| null |
Why did House Targaryen leave Valyria?
|
Twelve years before Valyria was destroyed by the mysterious "Doom," Aenar Targaryen's daughter, Daenys the Dreamer, had a prophetic vision of Valyria being destroyed, so he decided to take his family to Dragonstone.
|
|
null | false
| null |
Which of these terms are used in baseball and which are used in football: touchdown, home run, walk, first down, line drive, interception, strike out, offsides, holding, balk.
|
The baseball terms are: home run, walk, line drive, strike out, and balk. The football terms are: touchdown, first down, interception, offsides, and holding.
|
|
null | false
| null |
What are the three styles of Shakespeare?
|
Shakespeare's works were originally divided into three main styles or genres: Comedies, Tragedies and Histories.
|
|
null | false
| null |
Please classify each of the following animals as mammal, amphibian, or fish: elk, frog, dog, cat, newt, and human.
|
Newt and frog are amphibians. The rest of the animals are mammals. You did not provide any fish in your question.
|
|
null | false
| 421
|
Assume that we have an ML-model which implements the overall function h : R(I) → R(Y ), where I consists of all the input variables and Y is the single output variable. As mentioned, we also assume knowledge of a causal model M of the target domain (i.e., that part of the world which h is a model of). A first requirement is that such a causal model is consistent with the ML-model, meaning that they at least agree on how to classify all observations. A second requirement that is helpful given the conceptual nature of our analysis is that the endogenous variables I suffice to obtain deterministic causal knowledge of the target domain. Technically this means that there are only two kinds of endogenous variables: those which are determined directly by the unobserved exogenous variables, and those which are determined entirely by members of I. In future work both requirements can be loosened by adding probabilities to each. For the first, we can demand that both models are likely to agree on observations. For the second, we can add new exogenous variables and a probability distribution over them, which give us probabilistic causal models. These allow for probabilistic generalizations of all the definitions here presented.
either has a unique single exogenous parent U i ∈ U or no exogenous parents, and for all i ∈ R(I), y ∈ R(Y ):
As knowledge of the causal model is hard to come by many approaches to action-guiding explanations ignore it entirely and assume that the input variables are causally independent.
Definition 4 A causal model M that agrees with h satisfies Independence if for each I a , I b ∈ I, I a is not a parent of I b .
Obviously Independence rarely holds, and thus any account of action-guiding explanation that depends on it is limited. Although this limitation is recognized, this paper offers several results that formally characterize how extreme this limitation is. Simply put, under Independence a variety of very different causal notions become indistinguishable from each other. This holds for notions of sufficient explanation, counterfactual explanations, and various notions of actual causation.
Because of this limitation, work on action-guiding explanations in XAI has failed to take up the most relevant lesson that the literature on causation has to offer, namely that to give a causal explanation of an outcome is to give actual causes of that outcome. An additional reason for this oversight is that the precise relation between explanation and prediction has been the subject of much debate in the history of philosophy of science.
Prediction in its most natural application is a forward-looking notion, meaning one predicts an event before it takes place. Explanation on the other hand is a backward-looking notion, meaning that one explains an event after it has happened. Yet as many papers on XAI clearly illustrate, explanations about past events are often required precisely to inform predictions about future events. Therefore a suitable notion of causal explanation, and thus also of actual causation, needs to specify how it relates to predictions. Given the tumultuous history that this relation has in the philosophy literature, it has been duly neglected in the philosophical work on causation, thereby obscuring the importance of causation for the practical goals that XAI is concerned with. forms a notable exception!) 2 This paper corrects this by developing an account of causal explanation that shows both how it is connected to actual causation and how it can lead to action-guiding predictions.
Simply put, the goal of this paper is to upgrade the formalization of Woodward's influential philosophical account of causal explanation described below with the most recent insights from the causation literature, whilst also keeping track of the action-guiding demands that are prevalent in XAI.
Put differently, my idea is that one ought to be able to associate with any successful explanation a hypothetical or counterfactual experiment that shows us that and how manipulation of the factors mentioned in the explanation (the explanans, as philosophers call it) would be a way of manipulating or altering the phenomenon explained (the explanandum). Put in still another way, an explanation ought to be such that it can be used to answer what I call a what-if-things-had-been-different question: the explanation must enable us to see what sort of difference it would have made for the explanandum if the factors cited in the explanans had been different in various possible ways.
A crucial novel element of my account is the addition that a successful explanation must also be explicit about those factors that may not be manipulated for the explanation to hold, i.e., it must state which variables are to be safeguarded from interventions. Importantly, this is distinct from stating which variables must be held fixed at their actual values, for to hold variables fixed in fact means to intervene on them.
The following example (modified from) is helpful for illustrating the various purposes that explanations can serve.
Example 1 Consider a system for loan applications that is captured by a causal model such that
where Y is a binary variable representing whether the loan is granted, X 1 is the applicant's income, and X 2 is the applicant's savings. Further, assume that the applicant's savings are determined by their initial deposit X 3 and their income in the following manner: X 2 = 3/10 • X 1 + X 3 . It is also the case that people with high savings take out a safety deposit box (X 4 ) at the bank: X 4 = X 2 > 1, 000, 000.
Standard predictions in ML take place on the first rung of Pearl's causal ladder, namely that of observations. For example, an ML-model might pick up on the fact that all observed loan applicants who have a safety deposit box (X 4 = 1) also get a loan (Y = 1), and thus could predict that an applicant who has a safety deposit box will get a loan, i.e., it might learn a function so that h(x 1 , x 2 , x 3 , 1) = 1 for all values x 1 , x 2 , x 3 . In terms of the causal model, it is indeed the case that refer to as an anchor, which they interpret as a sufficient explanation of the outcome. Yet clearly such observational explanations are not a good guide towards action, for it would be absurd to recommend to someone to take out a safety deposit box so that their loan application is approved. This point can be brought out by making use of the second rung of the ladder, namely that of interventions: an applicant who takes out a safety deposit box as the result of being advised to do so will not necessarily get a loan:
In the same manner, interventions can be used to offer advice that is a good guide towards action, for example by telling an applicant that if they manage to obtain savings of 45, 001 then they are guaranteed to get a loan:
As mentioned, we also assume knowledge of a causal model M of the target domain (i.e., that part of the world which
h is a model of).
|
First paragraph of section 3: The term "target domain" is somewhat vague, what exactly needs to be known?
|
The target domain here simply refers to that part of the world that is being described by the variables and their interaction. I will make this explicit.
|
null | false
| null |
Name all the movies in the Harry Potter series
|
Here is the list of all the movies in the Harry Potter Series - Harry Potter and the Philosopher's Stone ,Harry Potter and the Chamber of Secrets, Harry Potter and the Prisoner of Azkaban, Harry Potter and the Goblet of Fire, Harry Potter and the Order of the Phoenix, Harry Potter and the Half-Blood Prince, Harry Potter and the Deathly Hallows - Part 1and Harry Potter and the Deathly Hallows - Part 2
|
|
null | false
| null |
The Cyber PHA methodology reconciles the process safety and cybersecurity approaches and allows IT, Operations and Engineering to collaborate in way that is already familiar to facility operations management and personnel. Modeled on the process safety PHA/HAZOP methodology, a cyber PHA enables cyber risks to be identified and analyzed in the same manner as any other process risk, and, because it can be conducted as a separate follow-on activity to a traditional HAZOP it can be used in both existing brownfield sites and newly constructed greenfield sites without unduly meddling with well established process safety processes.
|
what traditional process does Cyber PHA follow given this information?
|
HAZOP
|
null | false
| 76
|
We evaluate the generated supervised models for the two tasks, AEP and ASP, by splitting the train and testing instances. It is important to note that for the pairs INLINEFORM0 and the triple INLINEFORM1 , the news article INLINEFORM2 is referenced at time INLINEFORM3 by entity INLINEFORM4 , while the features take into account the entity profile at time INLINEFORM5 . This avoids any `overlapping' content between the news article and the entity page, which could affect the learning task of the functions INLINEFORM6 and INLINEFORM7 . Table TABREF47 shows the statistics of train and test instances. We learn the functions at year INLINEFORM8 and test on instances for the years greater than INLINEFORM9 . Please note that we do not show the performance for year 2014 as we do not have data for 2015 for evaluation.
We consider the following baselines for this task. • B1. The first baseline uses only the salience-based features by Dunietz and Gillick. • B2. The second baseline assigns the value relevant to a pair h n, ei , if and only if e appears in the title of n.
|
What baseline does the author consider for the article - entity location?
|
• B1. The first baseline uses only the salience-based features by Dunietz and Gillick. • B2. The second baseline assigns the value relevant to a pair h n, ei , if and only if e appears in the title of n.
|
1911.09483
| false
| null |
As shown in Table TABREF24, MUSE outperforms all previously models on En-De and En-Fr translation, including both state-of-the-art models of stand alone self-attention BIBREF0, BIBREF13, and convolutional models BIBREF11, BIBREF15, BIBREF10. This result shows that either self-attention or convolution alone is not enough for sequence to sequence learning. The proposed parallel multi-scale attention improves over them both on En-De and En-Fr.
Compared to Evolved Transformer BIBREF19 which is constructed by NAS and also mixes convolutions of different kernel size, MUSE achieves 2.2 BLEU gains in En-Fr translation.
As shown in Table TABREF24, MUSE outperforms all previously models on En-De and En-Fr translation, including both state-of-the-art models of stand alone self-attention BIBREF0, BIBREF13, and convolutional models BIBREF11, BIBREF15, BIBREF10. This result shows that either self-attention or convolution alone is not enough for sequence to sequence learning. The proposed parallel multi-scale attention improves over them both on En-De and En-Fr.
Compared to Evolved Transformer BIBREF19 which is constructed by NAS and also mixes convolutions of different kernel size, MUSE achieves 2.2 BLEU gains in En-Fr translation.
|
How big is improvement in performance over Transformers?
|
The answers are shown as follows:
* 2.2 BLEU gains
|
null | false
| null |
Does Bolivia have two capital cities?
|
The executive capital of Bolivia is La Paz. The constitutional capital is Sucre.
|
|
null | false
| null |
Identify which instrument is string or percussion: Shime-daiko, Chilador
|
Childador is string, Shime-daiko is percussion.
|
|
null | false
| null |
Are Air Fryers safe to use?
|
Air Fryers are really just small sized convection ovens (or OTGs) with a fan to circulate hot air more efficiently than in a large oven. This allows you to dehydrate the surface of foods really quickly and with very little oil.
As long as you don't char your food till it's visibly dark brown or black all over, air fried food is perfectly safe!
Convection baking has been used by human beings since the dawn of cooking, and air fryers (despite their name) use the exact same physics of hot air transferring heat to food.
And if you are someone looking to cut down calories from fat and also like crispy textures, this appliance is perfect for you. As always, focus on what you are ultimately eating rather than the appliance.
|
|
null | false
| null |
Provide a list of the best baseball players of all time
|
1. Babe Ruth
2. Ted Williams
3. Mickey Mantle
4. Tony Gwynn
5. Ken Griffey Jr
6. Cal Ripken Jr
7. Mike Trout
8. Hank Aaron
|
|
1911.12579
| false
| null |
The corpus is a collection of human language text BIBREF31 built with a specific purpose. However, the statistical analysis of the corpus provides quantitative, reusable data, and an opportunity to examine intuitions and ideas about language. Therefore, the corpus has great importance for the study of written language to examine the text. In fact, realizing the necessity of large text corpus for Sindhi, we started this research by collecting raw corpus from multiple web resource using web-scrappy framwork for extraction of news columns of daily Kawish and Awami Awaz Sindhi newspapers, Wikipedia dumps, short stories and sports news from Wichaar social blog, news from Focus Word press blog, historical writings, novels, stories, books from Sindh Salamat literary websites, novels, history and religious books from Sindhi Adabi Board and tweets regarding news and sports are collected from twitter.
In fact, realizing the necessity of large text corpus for Sindhi, we started this research by collecting raw corpus from multiple web resource using web-scrappy framwork for extraction of news columns of daily Kawish and Awami Awaz Sindhi newspapers, Wikipedia dumps, short stories and sports news from Wichaar social blog, news from Focus Word press blog, historical writings, novels, stories, books from Sindh Salamat literary websites, novels, history and religious books from Sindhi Adabi Board and tweets regarding news and sports are collected from twitter.
|
How is the data collected, which web resources were used?
|
The answers are shown as follows:
* daily Kawish and Awami Awaz Sindhi newspapers
* Wikipedia dumps
* short stories and sports news from Wichaar social blog
* news from Focus Word press blog
* historical writings, novels, stories, books from Sindh Salamat literary website
* novels, history and religious books from Sindhi Adabi Board
* tweets regarding news and sports are collected from twitter
|
null | false
| null |
What is District 13 in "The Hunger Games" known for developing?
|
District 13 is responsible for the development of nuclear weapons.
|
|
null | false
| null |
Since the eyes are believed to be expressive, the eyes portrayed in Maitum jars are given high regard in describing the dead. The eyes are classified to four major shapes: (1) almond, (2) ovaloid, (3) round and (4) rectangular.
|
Find the four eye shapes on Maitum anthropomorphic pottery and separate them with a comma
|
almond, ovaloid, round, rectangular
|
null | false
| null |
At the dawn as a social science, economics was defined and discussed at length as the study of production, distribution, and consumption of wealth by Jean-Baptiste Say in his Treatise on Political Economy or, The Production, Distribution, and Consumption of Wealth (1803). These three items are considered by the science only in relation to the increase or diminution of wealth, and not in reference to their processes of execution. Say's definition has prevailed up to our time, saved by substituting the word "wealth" for "goods and services" meaning that wealth may include non-material objects as well. One hundred and thirty years later, Lionel Robbins noticed that this definition no longer sufficed, because many economists were making theoretical and philosophical inroads in other areas of human activity. In his Essay on the Nature and Significance of Economic Science, he proposed a definition of economics as a study of a particular aspect of human behaviour, the one that falls under the influence of scarcity, which forces people to choose, allocate scarce resources to competing ends, and economize (seeking the greatest welfare while avoiding the wasting of scarce resources). For Robbins, the insufficiency was solved, and his definition allows us to proclaim, with an easy conscience, education economics, safety and security economics, health economics, war economics, and of course, production, distribution and consumption economics as valid subjects of the economic science." Citing Robbins: "Economics is the science which studies human behavior as a relationship between ends and scarce means which have alternative uses". After discussing it for decades, Robbins' definition became widely accepted by mainstream economists, and it has opened way into current textbooks. Although far from unanimous, most mainstream economists would accept some version of Robbins' definition, even though many have raised serious objections to the scope and method of economics, emanating from that definition. Due to the lack of strong consensus, and that production, distribution and consumption of goods and services is the prime area of study of economics, the old definition still stands in many quarters.
|
From the passage provided, extract the names and published work of the early economists. Identify the name and then the published work separated with a dash. Separate the name and published works pairs with a semi-colon.
|
Jean-Baptiste Say - Treatise on Political Economy or, The Production, Distribution, and Consumption of Wealth; Lionel Robbins - Essay on the Nature and Significance of Economic Science
|
null | false
| 477
|
Overall Time Complexity of Original Algorithm.****Figure A7: Time comparisons of GraphANGEL against baseline models on Amazon dataset in (a) and on
FB15k-237 dataset in (b).
|
What about the overall time complexity of the proposed method? In my view, the method could cost more time than other baselines, such as those purely embedding-based methods for link prediction.
|
We provide the overall time complexity analysis in Appendix 3.1. It is true that our method costs more time than those knowledge embedding-based methods such as TransE [1] and RotatE [2], as shown in Figure A7. However, we do note that the higher cost compared to traditional embedding-based methods also exist in other GNN-based methods such as R-GCN [3], as shown in Figure A7.
|
null | false
| null |
Carlos Andrés Pérez Rodríguez (27 October 1922 – 25 December 2010) also known as CAP and often referred to as El Gocho (due to his Andean origins), was a Venezuelan politician and the president of Venezuela from 12 March 1974 to 12 March 1979 and again from 2 February 1989 to 21 May 1993. He was one of the founders of Acción Democrática, the dominant political party in Venezuela during the second half of the twentieth century.
|
Who is Carlos Andres Perez?
|
Carlos Andrés Pérez Rodríguez (27 October 1922 – 25 December 2010) also known as CAP and often referred to as El Gocho (due to his Andean origins), was a Venezuelan politician and the president of Venezuela from 12 March 1974 to 12 March 1979 and again from 2 February 1989 to 21 May 1993. He was one of the founders of Acción Democrática, the dominant political party in Venezuela during the second half of the twentieth century.
His first presidency was known as the Saudi Venezuela due to its economic and social prosperity thanks to enormous income from petroleum exportation. However, his second presidency saw a continuation of the economic crisis of the 1980s, a series of social crises, widespread riots known as Caracazo and two coup attempts in 1992. In May 1993 he became the first Venezuelan president to be forced out of office by the Supreme Court on charges for the embezzlement of 250 million bolívars (roughly 2.7 million US dollars) belonging to a presidential discretionary fund, whose money was used to support the electoral process in Nicaragua and hire bodyguards for President Violeta Chamorro.
|
null | false
| null |
What is the minimal set of garden tools to start gardening?
|
For an outdoor garden, you only need a spade, clippers, watering can or hose, a rake and a shovel.
For an indoor garden, you will need only a spade and a watering can.
|
|
null | false
| null |
Identify which instrument is string or percussion: Tap shoe, Sarod
|
Sarod is string, Tap shoe is percussion.
|
|
null | false
| null |
Gerty Theresa Cori was a Czech-American biochemist who in 1947 was the third woman to win a Nobel Prize in science, and the first woman to be awarded the Nobel Prize in Physiology or Medicine, for her role in the "discovery of the course of the catalytic conversion of glycogen". Together with her husband Carl and Argentine physiologist Bernardo Houssay, Gerty Cori received the Nobel Prize in 1947 for the discovery of the mechanism by which glycogen—a starch made from glucose—is broken down in muscle tissue into lactic acid and then resynthesized in the body and stored as a source of energy (known as the Cori cycle). They also identified the important catalyzing compound, the Cori ester. The Coris were the third ever married couple to win the Nobel Prize. In 2004, both Gerty and Carl Cori were designated a National Historic Chemical Landmark in recognition of their work in clarifying carbohydrate metabolism.
|
Name some Nobel Prize winners
|
In 1947, biochemist Gerty Cori and her husband, Carl, received the Nobel Prize (along with their colleague physiologist Bernardo Houssay) for their work on the course of glycogen in the human body, now known as the Cori cycle.
|
null | false
| null |
What kind of weather do people prefer?
|
This is dependant on the type of person you are? The characteristics of the person, the location of the person, and the mood of the person. In general people prefer the Sun and comfortable amount of heat. The sun provides vitamin D. However during periods such as the Christmas period, most prefer the snow as it compliments the season. Also after warm summers, the winter is welcomed and see as a time to slow down and relax at home in a cosy envirnment.
|
|
null | false
| null |
What is the primary source of energy for heating homes in the Northeast United States.
|
Oil fired furnaces are predominant source of heat for homes in the Northeast, especially the New England states.
|
|
null | false
| null |
Which of the following numbers are used in numbered golf irons: 1, 2, 3, 4, 5, 6, 7, 8, 9, 10
|
2, 3, 4, 5, 6, 7, 8, 9
|
|
null | false
| null |
ICICI Bank Limited is an Indian multinational bank and financial services company headquartered in Mumbai. It offers a wide range of banking products and financial services for corporate and retail customers through a variety of delivery channels and specialized subsidiaries in the areas of investment banking, life, non-life insurance, venture capital and asset management.
This development finance institution has a network of 5,275 branches and 15,589 ATMs across India and has a presence in 17 countries. The bank has subsidiaries in the United Kingdom and Canada; branches in United States, Singapore, Bahrain, Hong Kong, Qatar, Oman, Dubai International Finance Centre, China and South Africa; as well as representative offices in United Arab Emirates, Bangladesh, Malaysia and Indonesia. The company's UK subsidiary has also established branches in Belgium and Germany.
|
From the passage identify the locations of the ICICI Bank branches and its subsidiaries outside India. Display the results in comma separated format.
|
United Kingdom, Canada, United States, Singapore, Bahrain, Hong Kong, Qatar, Oman, Dubai International Finance Centre, China, South Africa, United Arab Emirates, Bangladesh, Malaysia, Indonesia, Belgium, Germany
|
null | false
| null |
What is the most popular sport?
|
This depends on where you live, but football is one of the most popular sports in the world. In some countries, it is called soccer. Every 4 years countries assemble teams to see which is the best, and most countries have several leagues. Popular teams, such as real madrid, can garner fans from across the world. There are people who would even argue that football is more than just a sport.
|
|
null | false
| null |
How do you maintain good dental hygiene
|
Good dental hygiene is maintained by brushing and flossing your teeth everyday and eating low amounts of sugar.
|
|
null | false
| null |
Write a conversation between two long-time friends who run into one another five years after they initially drifted apart.
|
“Georgia, is that you?”
“Neha? My god, it has been ages! How are you.”
“I’m doing well, I just got back from Aspen. It was such a delight to be able to enjoy that, Roger’s parents were watching Jon, so it was nice to have some time to myself and catch up with some close friends. How are you?”
“I’m good, it’s been a really fun winter, our family from Canada was visiting and we did all the Holiday things, Gingerbread houses, Christmas tree decorating, cookies the whole shebang! Roger’s family is really into Christmas too, right Neha? Did you guys celebrate in Aspen?”
“Actually, Roger and I separated recently” Neha bites her lip as though to stop herself from crying. “Jon spent Christmas with Roger and his family this year. I went to Aspen with my parents and sister.”
Georgia looks aghast. “I’m so sorry, I thought if anyone would’ve made it, it would have been the two of you. Why didn’t you call?”
Neha looks up and thinks carefully about how to word it so she doesn’t come across as cross, although she is, she is very cross indeed. “You and I haven’t talked in ages, you weren’t exactly at the top of my list, you know? And the whole split happened while I was swamped at work, actually, you know, work was why we split up, and it’s just not been my way to reach out and reconnect.” Neha finishes and pauses.
Georgia looks hurt, but she understands. “That’s fair, I haven’t done a good job of keeping up. What do you mean you split up because of work?”
“I took a job, in a city he hated, and he didn’t want to uproot his life, and I didn’t want to miss out on this. I’m hopeful I’ll move back in a few years and things will go back to normal, but it’s a gamble.”
Georgia nods “Neha, I have to catch the next train, but please call, let’s get a glass of wine soon and really talk. And it’s his fault for hating New York.”
Neha smiles. She has a good feeling about this.
|
|
null | false
| null |
Name some of the bravolebrities from Southern Charm
|
Austen Kroll, Craig Conover, Kathryn Dennis, Leva Bonaparte, Madison LeCroy, Marcie Hobbs, Naomie Olindo, Olivia Flowers, Patricia Altschul, Shep Rose, Taylor Ann Green and Venita Aspen
|
|
1808.05077
| false
| null |
To evaluate the performance of the proposed approach, precision (1), recall (2), f-Measure (3), and prediction accuracy (4) have been used as a performance matrices. The experimental results are shown in Table 1, where it can be seen that autoencoders outperformed MLP and CNN outperformed autoencoders with the highest achieved accuracy of 82.6%. DISPLAYFORM0 DISPLAYFORM1
The experimental results are shown in Table 1, where it can be seen that autoencoders outperformed MLP and CNN outperformed autoencoders with the highest achieved accuracy of 82.6%.
|
Which deep learning model performed better?
|
The answers are shown as follows:
* autoencoders
|
null | false
| null |
What are some places to visit in Anchorage?
|
Anchorage has a great food scene. Restaurants and breweries include Moose's Tooth, Spenard Roadhouse, and 49th State Brewing. Cafes include Kaladi Brothers and Steamdot. Bakeries include Fire Island Rustic Bakeshop, and Wild Scoops is a great ice cream shop.
|
|
null | false
| 477
|
We systematically investigate the effect of three shapes of patterns used in GraphANGEL. For each dataset, we evaluate the effect of the patterns in 3-cycle and 4-cycle shapes by the performance of GraphANGEL 3−cycle and GraphANGEL 4−cycle . Since different shapes of patterns represent different composition logical rules, as shown in Table, patterns in Pair shape are the most general but include the least structure information, while those in 4-cycle shape are rich in the structure but less common. Hence, these patterns have their unique power in representing logics. Although it is hard to determine whether 3-cycle or 4-cycle shaped patterns is more powerful, as shown in Tables and A4, GraphANGEL with patterns in all shapes achieves the best performance. Besides the pattern type, we also investigate how the number of sampled subgraphs affects the performance. Taking Amazon and Douban Book datasets as examples, we show the performance of GraphANGEL under different K and Q in terms of ACC and AUC in Figure. One explanation is that the subgraphs following target patterns are constricted within the neighborhood of source and target nodes, the number of which is much smaller than subgraphs following supporting and refuting patterns.
Table 5: Result comparisons with baselines on generalization setting by randomly removing 20% relations. See Appendix A6.2 for full version and Appendix A6.2 for results of dropping 5%, 10%, 15%. The numbers in brackets show the descent degree.
|
Did you consider cases where there are rare relations?
|
We show the experimental results for GraphANGEL and baselines on the 20% least frequent relations in Table 5, where each model is trained with all the relations but tested only with the 20% least frequent relations. Results show that our method can significantly outperform the existing methods in terms of MRR, Hit@1, Hit@3, Hit@10.
|
null | false
| null |
The Western Iranian languages themselves are divided into two subgroups: Southwestern Iranian languages, of which Persian is the most widely spoken, and Northwestern Iranian languages, of which Kurdish and Balochi are the most widely spoken
|
Given this paragraph on the Persian language, which are the 2 sub groups of the Western Iranian language?
|
Southwestern and Northwestern Iranian language
|
null | false
| null |
Why is it a good idea to walk every day?
|
Walking is a good exercise for burning calories without causing any damage to joints. People of any age can do it and a few miles a day can result in positive effects on cardiovascular health and general fitness.
|
|
null | false
| null |
Why would a person from Switzerland immigrate to the California?
|
The warm climate, the nice weather, the proximity to the ocean in California are all things that Switzerland misses. People are usually more relaxed and friendlier to strangers than in Switzerland. In California has a huge variety of restaurants and grocery stores from all over the world.
But of course a person from Switzerland missed the mountains, excellent public transportation, cheese and chocolate.
|
|
null | false
| null |
What is a homonym?
|
Homonyms are words that are pronounced the same but have different meanings. Although the words sound the same when pronounced, they are not spelled the same. An example is the word deer, also spelled dear. These words have very different meanings in the english language. Homonyms can be found in all languages around the world.
|
|
null | false
| null |
What are some of the most popular tourist attractions in Rome?
|
Rome, Italy is of the most popular travel destinations in the world. People travel to Rome in particular to visit the Roman Colosseum and the Forum, which are ancient ruins. Tourists often make a quick visit to Vatican City to visit the Vatican Museum, which houses world renowned pieces of art. The Trevi Fountain and the Piazza Venezia are also quite popular.
|
|
null | false
| null |
What are examples of plants that can successfully grow in the tundra?
|
Arctic Moss, Arctic Poppy, Lichens, Labrador Tea, Arctic Willow.
|
|
null | false
| null |
Federalism is a combined and compound mode of government that combines a general government (the central or "federal" government) with regional governments (provincial, state, cantonal, territorial, or other sub-unit governments) in a single political system, dividing the powers between the two. Federalism in the modern era was first adopted in the unions of states during the Old Swiss Confederacy.Federalism differs from confederalism, in which the general level of government is subordinate to the regional level, and from devolution within a unitary state, in which the regional level of government is subordinate to the general level. It represents the central form in the pathway of regional integration or separation, bounded on the less integrated side by confederalism and on the more integrated side by devolution within a unitary state.
|
Provide some of the examples of regional governments.
|
provincial, state, cantonal, territorial, or other sub-unit governments
|
null | false
| null |
What do people paint?
|
Painting is an activity that is for everyone. Artist paint as a career but regular people also paint because it can be therapeutic, relaxing and a great way to relieve stress. You don't even have to be good at painting to get all the benefits that come from painting. Painting is great way to learn to appreciate what is in front of you and what you have created and can even give you a better outlook on life and improve your own spirituality.
|
|
null | false
| 28
|
The task of generating natural language descriptions of structured data (such as tables) BIBREF2 , BIBREF3 , BIBREF4 has seen a growth in interest with the rise of sequence to sequence models that provide an easy way of encoding tables and generating text from them BIBREF0 , BIBREF1 , BIBREF5 , BIBREF6 .
For text generation tasks, the only gold standard metric is to show the output to humans for judging its quality, but this is too expensive to apply repeatedly anytime small modifications are made to a system. Hence, automatic metrics that compare the generated text to one or more reference texts are routinely used to compare models BIBREF7 . For table-to-text generation, automatic evaluation has largely relied on BLEU BIBREF8 and ROUGE BIBREF9 . The underlying assumption behind these metrics is that the reference text is gold-standard, i.e., it is the ideal target text that a system should generate. In practice, however, when datasets are collected automatically and heuristically, the reference texts are often not ideal. Figure FIGREF2 shows an example from the WikiBio dataset BIBREF0 . Here the reference contains extra information which no system can be expected to produce given only the associated table. We call such reference texts divergent from the table.
We show that existing automatic metrics, including BLEU, correlate poorly with human judgments when the evaluation sets contain divergent references (§ SECREF36 ). For many table-to-text generation tasks, the tables themselves are in a pseudo-natural language format (e.g., WikiBio, WebNLG BIBREF6 , and E2E-NLG BIBREF10 ). In such cases we propose to compare the generated text to the underlying table as well to improve evaluation. We develop a new metric, PARENT (Precision And Recall of Entailed N-grams from the Table) (§ SECREF3 ). When computing precision, PARENT effectively uses a union of the reference and the table, to reward correct information missing from the reference. When computing recall, it uses an intersection of the reference and the table, to ignore extra incorrect information in the reference. The union and intersection are computed with the help of an entailment model to decide if a text n-gram is entailed by the table. We show that this method is more effective than using the table as an additional reference. Our main contributions are:
We develop a new metric, PARENT (Precision And Recall of Entailed Ngrams from the Table) (§3).
|
What metric do they propose?
|
PARENT.
|
null | false
| null |
Kafka was originally developed at LinkedIn, and was subsequently open sourced in early 2011. Jay Kreps, Neha Narkhede and Jun Rao helped co-create Kafka. Graduation from the Apache Incubator occurred on 23 October 2012. Jay Kreps chose to name the software after the author Franz Kafka because it is "a system optimized for writing", and he liked Kafka's work.
|
From the passage provided find the names of the creators of Kafka. Separate them with a comma.
|
Jay Kreps, Neha Narkhede, Jun Rao
|
null | false
| 72
|
We tested the proposed UTCNN on two different datasets: FBFans and CreateDebate. FBFans is a privately-owned, single-topic, Chinese, unbalanced, social media dataset, and CreateDebate is a public, multiple-topic, English, balanced, forum dataset. Results using these two datasets show the applicability and superiority for different topics, languages, data distributions, and platforms.
The FBFans dataset contains data from anti-nuclear-power Chinese Facebook fan groups from September 2013 to August 2014, including posts and their author and liker IDs. There are a total of 2,496 authors, 505,137 likers, 33,686 commenters, and 505,412 unique users. Two annotators were asked to take into account only the post content to label the stance of the posts in the whole dataset as supportive, neutral, or unsupportive (hereafter denoted as Sup, Neu, and Uns). Sup/Uns posts were those in support of or against anti-reconstruction; Neu posts were those evincing a neutral standpoint on the topic, or were irrelevant. Raw agreement between annotators is 0.91, indicating high agreement. Specifically, Cohen’s Kappa for Neu and not Neu labeling is 0.58 (moderate), and for Sup or Uns labeling is 0.84 (almost perfect). Posts with inconsistent labels were filtered out, and the development and testing sets were randomly selected from what was left. Posts in the development and testing sets involved at least one user who appeared in the training set. The number of posts for each stance is shown on the left-hand side of Table TABREF12 . About twenty percent of the posts were labeled with a stance, and the number of supportive (Sup) posts was much larger than that of the unsupportive (Uns) ones: this is thus highly skewed data, which complicates stance classification. On average, 161.1 users were involved in one post. The maximum was 23,297 and the minimum was one (the author). For comments, on average there were 3 comments per post. The maximum was 1,092 and the minimum was zero.
To test whether the assumption of this paper – posts attract users who hold the same stance to like them – is reliable, we examine the likes from authors of different stances. Posts in FBFans dataset are used for this analysis. We calculate the like statistics of each distinct author from these 32,595 posts. As the numbers of authors in the Sup, Neu and Uns stances are largely imbalanced, these numbers are normalized by the number of users of each stance. Table TABREF13 shows the results. Posts with stances (i.e., not neutral) attract users of the same stance. Neutral posts also attract both supportive and neutral users, like what we observe in supportive posts, but just the neutral posts can attract even more neutral likers. These results do suggest that users prefer posts of the same stance, or at least posts of no obvious stance which might cause annoyance when reading, and hence support the user modeling in our approach.
The CreateDebate dataset was collected from an English online debate forum discussing four topics: abortion (ABO), gay rights (GAY), Obama (OBA), and marijuana (MAR). The posts are annotated as for (F) and against (A). Replies to posts in this dataset are also labeled with stance and hence use the same data format as posts. The labeling results are shown in the right-hand side of Table TABREF12 . We observe that the dataset is more balanced than the FBFans dataset. In addition, there are 977 unique users in the dataset. To compare with Hasan and Ng's work, we conducted five-fold cross-validation and present the annotation results as the average number of all folds BIBREF9 , BIBREF5 .
The FBFans dataset has more integrated functions than the CreateDebate dataset; thus our model can utilize all linguistic and extra-linguistic features. For the CreateDebate dataset, on the other hand, the like and comment features are not available (as there is a stance label for each reply, replies are evaluated as posts as other previous work) but we still implemented our model using the content, author, and topic information.
The FBFans dataset has more integrated functions than the CreateDebate dataset; thus our model can utilize all linguistic and extra-linguistic features.
|
Whose integrated functions are stronger? FBFans dataset or CreateDebate dataset?
|
The FBFans dataset.
|
null | false
| null |
What are some common bases for tacos?
|
Tacos are typically built on top of a soft wheat tortilla, soft corn tortilla, hard taco shell, lettuce, or a baked taco shell.
|
|
1910.05154
| true
| null |
The experiment settings from this paper and evaluation protocol for the Mboshi corpus (Boundary F-scores using the ZRC speech reference) are the same from BIBREF8. Table presents the results for bilingual UWS and multilingual leveraging. For the former, we reach our best result by using as aligned information the French, the original aligned language for this dataset. Languages closely related to French (Spanish and Portuguese) ranked better, while our worst result used German. English also performs notably well in our experiments. We believe this is due to the statistics features of the resulting text. We observe in Table that the English portion of the dataset contains the smallest vocabulary among all languages. Since we train our systems in very low-resource settings, vocabulary-related features can impact greatly the system's capacity to language-model, and consequently the final quality of the produced alignments. Even in high-resource settings, it was already attested that some languages are more difficult to model than others BIBREF9.
The experiment settings from this paper and evaluation protocol for the Mboshi corpus (Boundary F-scores using the ZRC speech reference) are the same from BIBREF8. Table presents the results for bilingual UWS and multilingual leveraging. For the former, we reach our best result by using as aligned information the French, the original aligned language for this dataset. Languages closely related to French (Spanish and Portuguese) ranked better, while our worst result used German. English also performs notably well in our experiments. We believe this is due to the statistics features of the resulting text. We observe in Table that the English portion of the dataset contains the smallest vocabulary among all languages. Since we train our systems in very low-resource settings, vocabulary-related features can impact greatly the system's capacity to language-model, and consequently the final quality of the produced alignments. Even in high-resource settings, it was already attested that some languages are more difficult to model than others BIBREF9.
|
Is the model evaluated against any baseline?
|
No.
|
null | false
| null |
In the series A Song of Ice and Fire, who is the founder of House Crakehall?
|
Crake the Boarkiller
|
|
null | false
| null |
Why do some countries outperform other countries in certain sports despite having a smaller population?
|
A country's success in a given sport at the international level will depend on a number of factors other than population alone. While a country's population is a good identifier for the total selection pool for high performing athletes to come from, it is one of many factors. A country's founding history, climate zone, and GDP are other notable factors in sport performance.
A nations history plays factor as many sports are baked in centuries of tradition. For example, Cricket, which is believed to have originated in England is most popular in countries that were originally British Colonies. Long traditions such as this set specific sports into the culture of a nation, increasing the percentage of the athlete pool that it can pull from and can increase the level at which that nation is competitive.
Climate zone can also influence the competitiveness of a country in given sports. Norway, fore example, earned the most olympic medals of any country in the 2022 olympic games with a total of 37 medals. In the summer games of 2020 (held in 2021), Norway ranked 20th overall with just 8 medals. Norways Climate and Topography have led winter sports to be extremely popular, and Norway has over time developed elite programs in cross country/nordic skiing, speed skating, curling, biathlon, and ski jumping.
A nations wealth will also influence performance in sport at the international level. This occurs predominantly for two reasons. First is that more money enables larger budgets on equipment, coaching, player development, participation in international circuits, data insights, medical staff etc. that over time create a performance separation from those who can not afford these things. Second is access to sports, many sports are quite expensive some examples include snow sports, gymnastics, swimming, hockey, and polo. Nations which can not afford to provide the facilities or support to develop athletes will often fail to draw a large talent pool and not be competitive at the international level.
|
|
null | false
| null |
The Heart Institute, University of São Paulo (Instituto do Coração da Universidade de São Paulo, or InCor) is one of the clinical institutes of the central University's teaching hospital (Hospital das Clínicas da Universidade de São Paulo) and recognized as one of the world's academic and clinical excellence centers in cardiology, cardiovascular medicine and cardiovascular surgery.
The Heart Institute was founded in 1963 (60 years ago) by the noted cardiovascular surgeon and professor of the Faculty of Medicine of the University of São Paulo, Dr. Euryclides de Jesus Zerbini.
|
Who founded The Heart Institute (InCor) in São Paulo?
|
The Heart Institute (InCor) was founded by Dr. Euryclides de Jesus Zerbini.
|
null | false
| null |
What do Catholics believe?
|
"Our profession of faith begins with God, for God is the First and the Last, the beginning and the end of everything. The Credo begins with God the Father, for the Father is the first divine person of the Most Holy Trinity; our Creed begins with the creation of heaven and earth, for creation is the beginning and the foundation of all God's works."
---the Catechism of the Catholic Church, no. 198
Catholic belief is succinctly expressed in the profession of faith or credo called the Nicene Creed:
The Nicene Creed
I believe in one God,
the Father almighty,
maker of heaven and earth, of all things visible and invisible.
I believe in one Lord Jesus Christ, the Only Begotten Son of God, born of the Father before all ages. God from God, Light from Light, true God from true God, begotten, not made, consubstantial with the Father; through him all things were made. For us men and for our salvation he came down from heaven, and by the Holy Spirit was incarnate of the Virgin Mary, and became man. For our sake he was crucified under Pontius Pilate, he suffered death and was buried, and rose again on the third day in accordance with the Scriptures. He ascended into heaven and is seated at the right hand of the Father. He will come again in glory to judge the living and the dead and his kingdom will have no end.
I believe in the Holy Spirit, the Lord, the giver of life, who proceeds from the Father and the Son, who with the Father and the Son is adored and glorified, who has spoken through the prophets.
I believe in one, holy, catholic and apostolic Church. I confess one Baptism for the forgiveness of sins and I look forward to the resurrection of the dead and the life of the world to come. Amen.
|
|
null | false
| null |
Anaplasmosis is a tick-borne disease affecting ruminants, dogs, and horses, and is caused by Anaplasma bacteria. Anaplasmosis is an infectious but not contagious disease. Anaplasmosis can be transmitted through mechanical and biological vector processes. Anaplasmosis can also be referred to as "yellow bag" or "yellow fever" because the infected animal can develop a jaundiced look. Other signs of infection include weight loss, diarrhea, paleness of the skin, aggressive behavior, and high fever.
Many different tick species can carry the bacteria that cause anaplasmosis. The two major bacterial pathogens are Anaplasma marginale and Anaplasma phagocytophilum. These microorganisms are Gram-negative, and infect red blood cells. Once the host is infected with anaplasmosis, the immune system will try to fight off and kill the infected red blood cells, but will also kill healthy red blood cells. The Anaplasma sparouinense species is responsible for a rare zoonosis, the Sparouine anaplasmosis, detected only in French Guiana, South America. This disease was described from a clandestine gold miner working deep in rainforest. Infection of his red blood cells led to a severe deterioration of his health and required his hospitalization. Molecular typing showed that Anaplasma sparouinense is distinct to all known species and more genetically related to recently described Anaplasma species causing infections in rainforest wild fauna of Brazil.
While there are no current live or inactivated vaccines effective for all strains of A. marginale approved by the USDA for anaplasmosis, there are other means of prevention. Tick and fly control for herds of ruminants can be effective but also labor intensive. Chemical methods can also be used, including sanitizing surgical equipment after each use. Tetracycline drugs are the most common treatment for anaplasmosis, and can provide the animal with immunity for a period of time. The disease is more common in the South and West parts of the United States, but is no longer considered a major problem since the use of tetracycline drugs.
|
What are the two major bacterial pathogens of Anaplasma bacteria?
|
The two major bacterial pathogens are Anaplasma marginale and Anaplasma phagocytophilum. These microorganisms are Gram-negative, and infect red blood cells.
|
null | false
| null |
Campanula cervicaria, also known as the bristly bellflower, is a species of flowering plant in the bellflower family, Campanulaceae. Native to Scandinavia and central Europe, C. cervicaria has also become naturalised in parts of the U.S. state of Minnesota. Its natural habitat is woodland edges, hillside meadows, dry meadows and banks, and it also flourishes where soil has been disturbed, for example after slash-and-burn or forest clearance. The plant has a hairy stem, with light blue flowers about 1 to 2 centimetres (0.4 to 0.8 in) long and grouped together. This C. cervicaria flower was photographed in Keila, Estonia.
|
Where can I find a bristly bellflower in the wild?
|
The bristly bellflower (Campanula cervicaria) is native to Scandinavia and Central Europe. You can find it in hillside meadows and woodland edges. It flourishes in places where soil has been upturned.
|
null | false
| 315
|
Pre-trained language models such as BERT BIBREF0 have significantly improved the accuracy of various language processing tasks. However, we cannot apply BERT to language generation tasks as is because its model structure is not suitable for language generation. Several pre-trained seq-to-seq models for language generation BIBREF1, BIBREF2 based on an encoder-decoder Transformer model, which is a standard model for language generation, have recently been proposed. These models have achieved blackstate-of-the-art results in various language generation tasks, including abstractive summarization.
However, when generating a summary, it is essential to correctly predict which part of the source text should be included in the summary. Some previous studies without pre-training have examined combining extractive summarization with abstractive summarization BIBREF3, BIBREF4. Although pre-trained seq-to-seq models have achieved higher accuracy compared to previous models, it is not clear whether modeling “Which part of the source text is important?” can be learned through pre-training.
blackThe purpose of this study is to clarify the blackeffectiveness of combining saliency models that identify the important part of the source text with a pre-trained seq-to-seq model in the abstractive summarization task. Our main contributions are as follows:
We investigated nine combinations of pre-trained seq-to-seq and token-level saliency models, where the saliency models share the parameters with the encoder of the seq-to-seq model or extract important tokens independently of the encoder.
We proposed a new combination model, the conditional summarization model with important tokens (CIT), in which a token sequence extracted by a saliency model is explicitly given to a seq-to-seq model as an additional input text.
We evaluated the combination models on the CNN/DM BIBREF5 and XSum BIBREF6 datasets. Our CIT model outperformed a simple fine-tuned model in terms of ROUGE scores on both datasets.
We proposed a new combination model, the conditional summarization model with important tokens (CIT), in which a token sequence extracted by a saliency model is explicitly given to a seq-to-seq model as an additional input text.
|
What is new combination model proposed by authors?
|
The authors proposed a new combination model, the conditional summarization model with important tokens (CIT), in which a token sequence extracted by a saliency model is explicitly given to a seq-to-seq model as an additional input text.
|
null | false
| 314
|
In information retrieval (IR), queries and documents are typically represented by term vectors where each term is a content word and weighted by tf-idf, i.e. the product of the term frequency and the inverse document frequency, or other weighting schemes BIBREF0 . The similarity of a query and a document is then determined as a dot product or cosine similarity. Although this works reasonably, the traditional IR scheme often fails to find relevant documents when synonymous or polysemous words are used in a dataset, e.g. a document including only “neoplasm" cannot be found when the word “cancer" is used in a query. One solution of this problem is to use query expansion BIBREF1 , BIBREF2 , BIBREF3 , BIBREF4 or dictionaries, but these alternatives still depend on the same philosophy, i.e. queries and documents should share exactly the same words.
While the term vector model computes similarities in a sparse and high-dimensional space, the semantic analysis methods such as latent semantic analysis (LSA) BIBREF5 , BIBREF6 and latent Dirichlet allocation (LDA) BIBREF7 learn dense vector representations in a low-dimensional space. These methods choose a vector embedding for each term and estimate a similarity between terms by taking an inner product of their corresponding embeddings BIBREF8 . Since the similarity is calculated in a latent (semantic) space based on context, the semantic analysis approaches do not require having common words between a query and documents. However, it has been shown that LSA and LDA methods do not produce superior results in various IR tasks BIBREF9 , BIBREF10 , BIBREF11 and the classic ranking method, BM25 BIBREF12 , usually outperforms those methods in document ranking BIBREF13 , BIBREF14 .
Neural word embedding BIBREF15 , BIBREF16 is similar to the semantic analysis methods described above. It learns low-dimensional word vectors from text, but while LSA and LDA utilize co-occurrences of words, neural word embedding learns word vectors to predict context words BIBREF10 . Moreover, training of semantic vectors is derived from neural networks. Both co-occurrence and neural word embedding approaches have been used for lexical semantic tasks such as semantic relatedness (e.g. king and queen), synonym detection (e.g. cancer and carcinoma) and concept categorization (e.g. banana and pineapple belong to fruits) BIBREF10 , BIBREF17 . But, Baroni et al. Baroni2014 showed that neural word embedding approaches generally performed better on such tasks with less effort required for parameter optimization. The neural word embedding models have also gained popularity in recent years due to their high performance in NLP tasks BIBREF18 .
Here we present a query-document similarity measure using a neural word embedding approach. This work is particularly motivated by the Word Mover's Distance BIBREF19 . Unlike the common similarity measure taking query/document centroids of word embeddings, the proposed method evaluates a distance between individual words from a query and a document. Our first experiment was performed on the TREC 2006 and 2007 Genomics benchmark sets BIBREF20 , BIBREF21 , and the experimental results showed that our approach was better than BM25 ranking. This was solely based on matching queries and documents by the semantic measure and no other feature was used for ranking documents.
In general, conventional ranking models (e.g. BM25) rely on a manually designed ranking function and require heuristic optimization for parameters BIBREF22 , BIBREF23 . In the age of information explosion, this one-size-fits-all solution is no longer adequate. For instance, it is well known that links to a web page are an important source of information in web document search BIBREF24 , hence using the link information as well as the relevance between a query and a document is crucial for better ranking. In this regard, learning to rank BIBREF22 has drawn much attention as a scheme to learn how to combine diverse features. Given feature vectors of documents and their relevance levels, a learning to rank approach learns an optimal way of weighting and combining multiple features.
We argue that the single scores (or features) produced by BM25 and our proposed semantic measure complement each other, thus merging these two has a synergistic effect. To confirm this, we measured the impact on document ranking by combining BM25 and semantic scores using the learning to rank approach, LamdaMART BIBREF25 , BIBREF26 . Trained on PubMed user queries and their click-through data, we evaluated the search performance based on the most highly ranked 20 documents. As a result, we found that using our semantic measure further improved the performance of BM25.
Taken together, we make the following important contributions in this work. First, to the best of our knowledge, this work represents the first investigation of query-document similarity for information retrieval using the recently proposed Word Mover's Distance. Second, we modify the original Word Mover's Distance algorithm so that it is computationally less expensive and thus more practical and scalable for real-world search scenarios (e.g. biomedical literature search). Third, we measure the actual impact of neural word embeddings in PubMed by utilizing user queries and relevance information derived from click-through data. Finally, on TREC and PubMed datasets, our proposed method achieves stronger performance than BM25.
This was solely based on matching queries and documents by the semantic measure and no other feature was used for ranking documents.
|
Was the other feature used for ranking documents?
|
No, it wasn't.
|
null | false
| null |
How do we get rainbows?
|
The electromagnetic spectrum is made of light with many different wavelengths consisting of a band of colors that include red, orange, yellow, green, blue, indigo and violet. White light is how our eyes perceive all the colors of the rainbow mixed together. Sunlight appears white. When sunlight hits a rain droplet, some of the light is reflected. They are reflected at a different angles separating the spectrum and producing a rainbow.
|
|
null | false
| 309
|
We train our MTNet with Adam optimizer for 20 epochs. We set batch size as 8, and weight decay as 1e-5, we adopt 3 branched layers in MTNet. Detailed configuration is shown in Table TABREF21 . We use ReLU in shared layers, and relu6 in branched layers to prevent information loss. Our proposed MTNet achieves 0.2250 on MAE and 0.3216 on RMSE, respectively.
We compare MTNet with other mainstream regression algorithms BIBREF14 (linear regression, KNN, SVR, Random Forest and MLP). The architecture of MLP is 15-16-8-8-1, where each number represents the number of neurons in each layer. We try three kinds of kernels (RBF kernel, linear kernel, and poly kernel) with SVR in our experiments for fair comparison.
The results are listed in Table TABREF37 . Our method achieves the best performance in contrast to the compared baseline regressors.
We compare MTNet with other mainstream regression algorithms [1] (linear regression, KNN, SVR, Random Forest and MLP).
|
What are their baselines?
|
Linear regression, KNN, SVR, Random Forest and MLP.
|
null | false
| 100
|
Understanding what a question is asking is one of the first steps that humans use to work towards an answer. In the context of question answering, question classification allows automated systems to intelligently target their inference systems to domain-specific solvers capable of addressing specific kinds of questions and problem solving methods with high confidence and answer accuracy BIBREF0 , BIBREF1 .
To date, question classification has primarily been studied in the context of open-domain TREC questions BIBREF2 , with smaller recent datasets available in the biomedical BIBREF3 , BIBREF4 and education BIBREF5 domains. The open-domain TREC question corpus is a set of 5,952 short factoid questions paired with a taxonomy developed by Li and Roth BIBREF6 that includes 6 coarse answer types (such as entities, locations, and numbers), and 50 fine-grained types (e.g. specific kinds of entities, such as animals or vehicles). While a wide variety of syntactic, semantic, and other features and classification methods have been applied to this task, culminating in near-perfect classification performance BIBREF7 , recent work has demonstrated that QC methods developed on TREC questions generally fail to transfer to datasets with more complex questions such as those in the biomedical domain BIBREF3 , likely due in part to the simplicity and syntactic regularity of the questions, and the ability for simpler term-frequency models to achieve near-ceiling performance BIBREF8 . In this work we explore question classification in the context of multiple choice science exams. Standardized science exams have been proposed as a challenge task for question answering BIBREF9 , as most questions contain a variety of challenging inference problems BIBREF10 , BIBREF11 , require detailed scientific and common-sense knowledge to answer and explain the reasoning behind those answers BIBREF12 , and questions are often embedded in complex examples or other distractors. Question classification taxonomies and annotation are difficult and expensive to generate, and because of the unavailability of this data, to date most models for science questions use one or a small number of generic solvers that perform little or no question decomposition BIBREF13 , BIBREF14 , BIBREF15 , BIBREF16 , BIBREF17 . Our long-term interest is in developing methods that intelligently target their inferences to generate both correct answers and compelling human-readable explanations for the reasoning behind those answers. The lack of targeted solving – using the same methods for inferring answers to spatial questions about planetary motion, chemical questions about photosynthesis, and electrical questions about circuit continuity – is a substantial barrier to increasing performance (see Figure FIGREF1 ).
To address this need for developing methods of targetted inference, this work makes the following contributions:
In this work we explore question classification in the context of multiple choice science exams.
|
What does the paper explore?
|
Question classification in the context of multiple choice science exams.
|
null | false
| null |
What is the best way to cook a steak?
|
Because tastes differs, there are no best way to cook a steak.
Here are some optimized ways to cook steak.
Fastest way to cook a steak - cast iron skillet. Pan sear room temperature steak in a host cast iron skillet; sear the steak on both sides; baste with melted butter / herbs until done to the desired doneness; rest/cut/serve.
Most consistent way to cook a steak - sou vide followed by flame torch the exterior. Season steak on both sides / edges; put into vacuum bag, vacuum bag, put under sous vide water bath at 130F for 1-4 hours; take it out of the bag, pat dry; use flame torch to char both sides; rest/cut/serve.
Most scientific way to cook a steak - cold sear. Put steak into non-sticking pan on high for 2 minutes; flip once and cook for another 2 minutes; reduce heat to medium low, flip and cook for another 2 minutes; flip and cook for another 2 minutes; keep on flipping / cooking for 2 minutes until reaching desired doneness; season; rest/cut/serve.
|
|
1710.00341
| false
| null |
This step consists of generating a query out of the claim and querying a search engine (here, we experiment with Google and Bing) in order to retrieve supporting documents. Rather than querying the search engine with the full claim (as on average, a claim is two sentences long), we generate a shorter query following the lessons highlighted in BIBREF0 .
We rank the words by means of tf-idf. We compute the idf values on a 2015 Wikipedia dump and the English Gigaword. BIBREF0 suggested that a good way to perform high-quality search is to only consider the verbs, the nouns and the adjectives in the claim; thus, we exclude all words in the claim that belong to other parts of speech. Moreover, claims often contain named entities (e.g., names of persons, locations, and organizations); hence, we augment the initial query with all the named entities from the claim's text. We use IBM's AlchemyAPI to identify named entities. Ultimately, we generate queries of 5–10 tokens, which we execute against a search engine. We then collect the snippets and the URLs in the results, skipping any result that points to a domain that is considered unreliable. Finally, if our query has returned no results, we iteratively relax it by dropping the final tokens one at a time.
This step consists of generating a query out of the claim and querying a search engine (here, we experiment with Google and Bing) in order to retrieve supporting documents. Rather than querying the search engine with the full claim (as on average, a claim is two sentences long), we generate a shorter query following the lessons highlighted in BIBREF0 .
We rank the words by means of tf-idf. We compute the idf values on a 2015 Wikipedia dump and the English Gigaword. BIBREF0 suggested that a good way to perform high-quality search is to only consider the verbs, the nouns and the adjectives in the claim; thus, we exclude all words in the claim that belong to other parts of speech. Moreover, claims often contain named entities (e.g., names of persons, locations, and organizations); hence, we augment the initial query with all the named entities from the claim's text. We use IBM's AlchemyAPI to identify named entities. Ultimately, we generate queries of 5–10 tokens, which we execute against a search engine. We then collect the snippets and the URLs in the results, skipping any result that points to a domain that is considered unreliable.
|
How are the potentially relevant text fragments identified?
|
Generate a query out of the claim and querying a search engine, rank the words by means of TF-IDF, use IBM's AlchemyAPI to identify named entities, generate queries of 5–10 tokens, which execute against a search engine, and collect the snippets and the URLs in the results, skipping any result that points to a domain that is considered unreliable.
|
null | false
| null |
Which of these are religions and which are countries: Christianity, Judaism, Norway, Canada, Islam, Egypt, China, and Iran.
|
The countries in this list are Norway, Canada, Egypt, China and Iran. The religions are Christian, Judaism, and Islam.
|
|
null | false
| null |
What are some horror movies from 1987?
|
There were many horror movies that were released in 1987. Here is a list of a few of them:
Evil Dead II
Hellraiser
Opera
Predator
Bad Taste
A Nightmare on Elm Street 3: Dream Warriors
|
|
null | false
| 77
|
We evaluate the corpus using Named Entity Recognition (NER), i. e., the task of finding mentions of concepts of interest in unstructured text. We focus on detecting cases, conditions, factors, findings and modifiers in case reports (cf. Section SECREF6). We approach this as a sequence labeling problem. Four systems were developed to offer comparable robust baselines.
The original documents are pre-processed (sentence splitting and tokenization with ScispaCy). We do not perform stop word removal or lower-casing of the tokens. The BIO labeling scheme is used to capture the order of tokens belonging to the same entity type and enable span-level detection of entities. Detection of nested and/or discontinuous entities is not supported. The annotated corpus is randomized and split in five folds using scikit-learn BIBREF9. Each fold has a train, test and dev split with the test split defined as .15% of the train split. This ensures comparability between the presented systems.
The original documents are pre-processed (sentence splitting and tokenization with ScispaCy[ https://github.eom/allenai/scispacy/releases/tag/v0.2.2 (Neumann et al., 2019) and the en_core_sci_md model.]).
|
Are the original documents pre-processed？
|
Yes.
|
null | false
| 49
|
To enable open-vocabulary translation, we segment words via byte-pair encoding (BPE) BIBREF1 . BPE, originally devised as a compression algorithm BIBREF6 , is adapted to word segmentation as follows:
First, each word in the training vocabulary is represented as a sequence of characters, plus an end-of-word symbol. All characters are added to the symbol vocabulary. Then, the most frequent symbol pair is identified, and all its occurrences are merged, producing a new symbol that is added to the vocabulary. The previous step is repeated until a set number of merge operations have been learned.
BPE starts from a character-level segmentation, but as we increase the number of merge operations, it becomes more and more different from a pure character-level model in that frequent character sequences, and even full words, are encoded as a single symbol. This allows for a trade-off between the size of the model vocabulary and the length of training sequences. The ordered list of merge operations, learned on the training set, can be applied to any text to segment words into subword units that are in-vocabulary in respect to the training set (except for unseen characters).
To increase consistency in the segmentation of the source and target text, we combine the source and target side of the training set for learning BPE. For each language pair, we learn 89500 merge operations.
To enable open-vocabulary translation, we segment words via byte-pair encoding (BPE)5 (Sennrich et al., 2016b). BPE, originally devised as a compression algorithm (Gage, 1994), is adapted to word segmentation as follows: First, each word in the training vocabulary is represented as a sequence of characters, plus an end-of-word symbol. All characters are added to the symbol vocabulary. Then, the most frequent symbol pair is identified, and all its occurrences are merged, producing a new symbol that is added to the vocabulary. The previous step is repeated until a set number of merge operations have been learned.
|
How does the BPE segment words?
|
First, each word in the training vocabulary is represented as a sequence of characters, plus an end-of-word symbol. All characters are added to the symbol vocabulary. Then, the most frequent symbol pair is identified, and all its occurrences are merged, producing a new symbol that is added to the vocabulary. The previous step is repeated until a set number of merge operations have been learned.
|
null | false
| null |
The Western States Trail Ride was first completed on foot by seven soldiers from Fort Riley, Kansas during the 1972 event, proving that the mountain course could be marched in less than two days. In 1974 Gordy Ainsleigh was the first to run it in under 24 hours. Ainsleigh had finished the Western States Trail Ride (Tevis Cup) in 1971 and 1972 on horseback, but in 1973 his new horse was pulled with lameness at the 29-mile checkpoint. In 1974, with the inspiration and encouragement of Drucilla Barner, the first woman to win the Tevis Cup and Secretary of the Western States Trail Foundation, Gordy joined the horses of the Western States Trail Ride to see if he could complete the course on foot in under 24 hours. Twenty-three hours and forty-two minutes later Gordy arrived in Auburn, proving that a runner could cover the Western States trail in a day. However, he did not run 100 miles. The course was at most 89 miles until 1980 when it was extended to 93.5 miles and in 1985 finally to 100 miles. Running 100 miles in a day was nothing new. About 1,000 runners worldwide had run or walked 100 miles in a day prior to 1974 on tracks, roads, and trails.
|
Given this paragraph, who was the first person to run the Western States 100 in under 24 hours?
|
In 1974 Gordy Ainsleigh was the first to run it in under 24 hours.
|
null | false
| null |
Identify which instrument is string or woodwind: Auto-harp, Saxophone
|
Auto-harp is string, Saxophone is woodwind.
|
|
null | false
| null |
List high quality binge worthy tv shows
|
Breaking Bad, Sons of Anarchy, Homeland, The Sopranos, Mind Hunter, Shameless
|
|
null | false
| null |
Johnson entered Hollywood and the film industry, becoming a star initially through his wrestling popularity and noted work ethic. Over his acting career, he became one of the highest paid and most successful actors in Hollywood. He began his acting career on television while wrestling. In his first television acting job, in 1999, he played his own father in an episode of That '70s Show called "That Wrestling Show". Nearly a year later, he appeared in the Star Trek: Voyager episode "Tsunkatse" as an alien wrestler who fought popular character Seven of Nine. While Johnson was away from WWE, the company continued to sell "The Rock" merchandise, and he continued to be featured prominently in the opening montages of their television shows.
Johnson began his theatrical career in The Mummy Returns (2001), The Scorpion King (2002), The Rundown (2003), and Walking Tall (2004). He played a supporting role in Be Cool (2005) - most notable for the meta element of playing a bodyguard that wants to become an actor - and was the primary antagonist in Doom (2005). He also had roles in Gridiron Gang (2006), Reno 911!: Miami (2007), and Southland Tales (2006). He played a cocky famous American football player in The Game Plan (2007) and Agent 23 in Get Smart (2008). He presented the Academy Award for Best Visual Effects at the 80th Academy Awards.
|
Name some of the movies where Johnson acted in and the years in which the movies were released.
|
The Mummy Returns (2001), The Scorpion King (2002), The Rundown (2003), Walking Tall (2004), Be Cool (2005), Doom (2005), Gridiron Gang (2006), Reno 911!: Miami (2007), Southland Tales (2006), The Game Plan (2007) and Get Smart (2008).
|
null | false
| null |
Electricity is the set of physical phenomena associated with the presence and motion of matter that has a property of electric charge. Electricity is related to magnetism, both being part of the phenomenon of electromagnetism, as described by Maxwell's equations. Various common phenomena are related to electricity, including lightning, static electricity, electric heating, electric discharges and many others.
The presence of either a positive or negative electric charge produces an electric field. The movement of electric charges is an electric current and produces a magnetic field. In most applications, a force acts on a charge with a magnitude given by Coulomb's law. Electric potential is typically measured in volts.
Electricity is at the heart of many modern technologies, being used for:
Electric power where electric current is used to energise equipment;
Electronics which deals with electrical circuits that involve active electrical components such as vacuum tubes, transistors, diodes and integrated circuits, and associated passive interconnection technologies.
Electrical phenomena have been studied since antiquity, though progress in theoretical understanding remained slow until the 17th and 18th centuries. The theory of electromagnetism was developed in the 19th century, and by the end of that century electricity was being put to industrial and residential use by electrical engineers. The rapid expansion in electrical technology at this time transformed industry and society, becoming a driving force for the Second Industrial Revolution. Electricity's extraordinary versatility means it can be put to an almost limitless set of applications which include transport, heating, lighting, communications, and computation. Electrical power is now the backbone of modern industrial society.
|
From the following passage, extract 3 common phenomena related to electricity.
|
3 common phenomena related to electricity are:
1. Lightning
2. Static electricity
3. Electric heating
|
null | false
| null |
In a bingo game, which number is represented by the phrase “two little ducks”?
|
22
|
|
null | false
| null |
What is the histroy behind electoral votes?
|
The Founding Fathers established the Electoral College in the Constitution, in part, as a compromise between the election of the President by a vote in Congress and election of the President by a popular vote of qualified citizens.
|
|
null | false
| 90
|
BERT stands for "Bidirectional Encoder Representations from Transformers" BIBREF1 is a contextual word embedding model. Given a sentence as an input, contextual embedding for the words are returned. The BERT model was designed so it can be fine tuned for 11 different tasks BIBREF1, including question answering tasks. For a question answering task, question and paragraph (context) are given as an input. A BERT standard is that question text and paragraph text are separated by a separator [Sep]. BERT question-answering fine tuning involves adding softmax layer. Softmax layer takes contextual word embeddings from BERT as input and learns to identity answer span present in the paragraph (context). This process is represented in Figure FIGREF4.
BERT was originally trained to perform tasks such as language model creation using masked words and next-sentence-prediction. In other words BERT weights are learned such that context is used in building the representation of the word, not just as a loss function to help learn a context-independent representation. For detailed understanding of BERT Architecture, please refer to the original BERT paper BIBREF1.
A ‘word embedding’ is a learned representation. It is represented in the form of vector where words that have the same meaning have a similar vector representation. Consider a word embedding model 'word2vec' BIBREF6 trained on a corpus. Word embeddings generated from the model are context independent that is, word embeddings are returned regardless of where the words appear in a sentence and regardless of e.g. the sentiment of the sentence. However, contextual word embedding models like BERT also takes context of the word into consideration.
BERT was originally trained to perform tasks such as language model creation using masked words and next-sentence-prediction.
|
What was BERT originally trained to do?
|
To perform tasks such as language model creation using masked words and next-sentence-prediction.
|
1709.00947
| false
| null |
For all these tests we computed a coverage metric. Our embeddings do not necessarily contain information for all the words contained in each of these tests. So, for all tests, we compute a coverage metric that measures the fraction of the gold-standard pairs that could actually be tested using the different embeddings produced. Then, for all the test pairs actually covered, we obtain the success metrics for each of the 3 tests by computing the ratio of pairs we were able to correctly classified as i) being distinct (cosine INLINEFORM0 0.7 or 0.8), ii) belonging to the same class (cosine INLINEFORM1 0.7 or 0.8), and iii) being equivalent (cosine INLINEFORM2 0.85 or 0.95).
For all these tests we computed a coverage metric. Our embeddings do not necessarily contain information for all the words contained in each of these tests. So, for all tests, we compute a coverage metric that measures the fraction of the gold-standard pairs that could actually be tested using the different embeddings produced.
Then, for all the test pairs actually covered, we obtain the success metrics for each of the 3 tests by computing the ratio of pairs we were able to correctly classified as i) being distinct (cosine INLINEFORM0 0.7 or 0.8), ii) belonging to the same class (cosine INLINEFORM1 0.7 or 0.8), and iii) being equivalent (cosine INLINEFORM2 0.85 or 0.95).
|
What intrinsic evaluation metrics are used?
|
The answers are shown as follows:
* coverage metric
* being distinct (cosine INLINEFORM0 0.7 or 0.8)
* belonging to the same class (cosine INLINEFORM1 0.7 or 0.8)
* being equivalent (cosine INLINEFORM2 0.85 or 0.95)
|
null | false
| null |
Which of the following are commonly used as boys names: Adam, Shirley, Hunter, Mary, Susan, Stephanie, John, Buster, Matthew, James, Johan, Sean, Steven, Ada
|
Adam, Hunter, John, Buster, Matthew, James, Johan, Sean, Steven
|
|
1812.07023
| true
| null |
The utterance is concatenated with a special symbol marking the end of the sequence. We initialize our word embeddings using 300-dimensional GloVe BIBREF30 and then fine-tune them during training.
|
Do they use pretrained word vectors for dialogue context embedding?
|
Yes.
|
null | false
| null |
List the 7 most recent US Speakers of the House (newest to oldest).
|
Kevin McCarthy
Nancy Pelosi
Paul Ryan
John Boehner
Nancy Pelosi
Dennis Hastert
Newt Gingrich
|
|
null | false
| null |
What town is the Simpsons set in?
|
Springfield
|
|
null | false
| 347
|
Zero anaphora is a discourse phenomenon, where pronouns can be omitted when they are pragmatically or grammatically inferable from intra- and inter-sentential context BIBREF0 . However, translating such implicit information (i.e. zero pronoun, ZP) poses various difficulties for machine translation (MT) in terms of completeness and correctness. Although neural models are getting better at learning representations, it is still difficult to implicitly learn complex ZPs in a general model. Actually, ZP prediction and translation need to not only understand the semantics or intentions of a single sentence, but also utilize its discourse-level context.
Two technological advances in the field of ZP and MT, have seen vast progress over the last decades, but they have been developed very much in isolation. Early studies BIBREF1 , BIBREF2 , BIBREF3 fed MT systems with the results of ZP prediction models, which are trained on a small-scale and non-homologous data compared to MT models. To narrow the data-level gap, Wang:2016:NAACL proposed an automatic method to annotate ZPs by utilizing the parallel corpus of MT. The homologous data for both ZP prediction and translation leads to significant improvements on translation performances for both statistical MT BIBREF4 and neural MT models BIBREF5 . However, such approaches still require external ZP prediction models, which have a low accuracy of 66%. The numerous errors of ZP prediction errors will be propagated to translation models, which leads to new translation problems. In addition, relying on external ZP prediction models in decoding makes these approaches unwieldy in practice, due to introducing more computation cost and pipeline complexity.
In this work, we try to further bridge the model-level gap by jointly modeling ZP prediction and translation. Joint learning has proven highly effective on alleviating the error propagation problem, such as joint parsing and translation BIBREF6 , as well as joint tokenization and translation BIBREF7 . Similarly, we expect that ZP prediction and translation could interact with each other: prediction offers more ZP information beyond 1-best result to translation and translation helps prediction resolve ambiguity. Specifically, we first cast ZP prediction as a sequence labeling task with a neural model, which is trained jointly with a standard neural machine translation (NMT) model in an end-to-end manner. We leverage the auto-annotated ZPs to supervise the learning of ZP prediction component, which releases the reliance on external ZP knowledge in decoding phase.
In addition, previous studies revealed that discourse-level information can better tackle ZP resolution, because around 23% of ZPs appear two or more sentences away from their antecedents BIBREF8 , BIBREF9 . Inspired by these findings, we exploit inter-sentential context to further improve ZP prediction and thus translation. Concretely, we employ hierarchical neural networks BIBREF10 , BIBREF11 to summarize the context of previous sentences in a text, which is integrated to the joint model for ZP prediction.
We validate the proposed approach on the widely-used data for ZP translation BIBREF5 , which consist of 2.15M Chinese–English sentence pairs. Experimental results show that the joint model indeed improves performances on both ZP prediction and translation. Incorporating discourse-level context further improves performances, and outperforms ther external ZP prediction model BIBREF5 by +2.29 BLEU points in translation and +11% in prediction accuracy. Experimental results on a further Japanese–English translation task show that our model consistently outperforms both the baseline and the external ZP prediction model, demonstrating the universality of the proposed approach.
The key contributions of this paper are:
Experimental results show that the joint model indeed improves performances on both ZP prediction and translation. Incorporating discourse-level context further improves performances, and outperforms ther external ZP prediction model by +2.29 BLEU points in translation and +11% in prediction accuracy.
|
The joint model improves performances on what according to their experimental results?
|
On both ZP prediction and translation.
|
null | false
| null |
What are the top 5 best movies ever made?
|
The top 5 best movies ever made are:
1. The Godfather (1972)
2. The Dark Knight (2008)
3. The Matrix (1999)
4. Star Wars (1977)
5. Jurassic Park (1993)
However, this largely depends on one's taste. There are some who claim that Weekend at Bernie's is the best movie ever made.
|
|
null | false
| null |
Tell me if the following music pieces are considered classical music, jazz music or pop music? The magic flute, Kind of blue, Wish you were here, Bach's Saint John Passion, Take five, We are the champions.
|
The music pieces The magic flute, Bach's Saint John Passion are considered classical music. The pieces Kind of blue and Take five are considered jazz. music and the pieces Wish you were here and We are the champions are considered pop music.
|
|
null | false
| null |
What is Edison famous for?
|
Inventing the light bulb
|
|
null | false
| null |
What are the best hikes near Portland Oregon.
|
There are many hikes near Portland Oregon, especially on the Columbia River Gorge. The Columbia River Gorge National Scenic Area is the best place to experience the beauty of the Pacific Northwest. The most popular hike in the gorge is Angel’s Rest. This hike is challenging but will only take you a few hours and the payoff is beautiful views of the gorge. Another very popular hike is at Multnomah Falls. This incredible waterfall is over 600 feet tall and makes for some impressive pictures from the bottom or top. You can make this hike as easy or difficult as you want. A lesser known but still incredible hike in the gorge is Hamilton Mountain. You will be treated to some incredible eastern gorge views if you continue on from the peak to the subsequent mountain saddle. When in doubt use the Alltrails app to find hikes in any area your travels take you.
|
|
null | false
| null |
How many planets make up the Solar System?
|
There are 8 planets in the solar system. Mercury is the smallest and closest to the Sun. Venus is the second closest to the Sun. Earth, our home planet, is the third planet closest to the Sun followed by Mars. After the asteroid belt is Jupiter, the fifth and biggest planet in the Solar System. Saturn, Uranus and Neptune are the last planets in the Solar System.
Originally the Solar System was said to have 9 planets until 2006 when the term "planet" was redefined to exclude "dwarf planets". This results in the exclusion of the ninth and farthest away planet, Pluto, reducing the count of planets in the Solar System to current count of 8.
|
|
null | false
| null |
Which is the best place to Kayak in Austin?
|
Lady Bird Lake in Austin downtown is revered to be the best place to go Kayaking. Paddlers can choose to head west from the Rowing Dock to take in nature views. But a small paddle down the river to the east will greet paddlers to magnificent downtown skyline views.
|
|
null | false
| null |
What countries made up Yugoslavia?
|
Yugoslavia was made up of 7 present day countries. These countries include Bosnia and Herzegovina, Croatia, Montenegro, North Macedonia, Serbia, Kosovo, and Slovenia.
|
|
null | false
| null |
Which of the following are not colors of the rainbow spectrum? Purple, Blue, Aqua, Orange, Black, Violet, Pink
|
The colors of the rainbow spectrum are typically referenced using the acronym ROYGBIV: Red, Orange, Yellow, Green, Blue, Indigo & Violet. From the prior list, the colors that are not part of the rainbow spectrum are Purple, Aqua, Black and Pink.
|
|
null | false
| null |
Please tell me if the following cities are in Germany, Austria or Switzerland: Munich, Vienna, Zurich, Dortmund, Lucerne, Bern, Stuttgart, Klagenfurt, Salzburg, Bremen, Geneva, Innsbruck
|
Germany: Munich, Dortmund, Stuttgart, Bremen
Austria: Vienna, Klagenfurt, Salzburg, Innsbruck
Switzerland: Zurich, Lucerne, Bern, Geneva
|
|
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.
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).
|
Which domains do those scientific articles cover?
|
Covering four computer science domains: natural language processing (NLP), information retrieval (IR), artificial intelligence (AI) and machine learning (ML).
|
null | false
| null |
What are Jindo dogs like?
|
Jindos are extremely loyal, and territorial. Unlike most dogs, Jindos do not play fetch, and rarely bark. They are also very independent, and some would say they march to the beat of their own drum. When meeting strangers they tend to be very reserved, but are highly affectionate to their owners. It is very hard to earn their trust, but once you do Jindos make loving companions.
|
|
null | false
| 192
|
Currently, probabilistic topic models are important tools for improving automatic text processing including information retrieval, text categorization, summarization, etc. Besides, they can be useful in supporting expert analysis of document collections, news flows, or large volumes of messages in social networks BIBREF0 , BIBREF1 , BIBREF2 . To facilitate this analysis, such approaches as automatic topic labeling and various visualization techniques have been proposed BIBREF1 , BIBREF3 .
Boyd-Graber et al. BIBREF4 indicate that to be understandable by humans, topics should be specific, coherent, and informative. Relationships between the topic components can be inferred. In BIBREF1 four topic visualization approaches are compared. The authors of the experiment concluded that manual topic labels include a considerable number of phrases; users prefer shorter labels with more general words and tend to incorporate phrases and more generic terminology when using more complex network graph. Blei and Lafferty BIBREF3 visualize topics with ngrams consisting of words mentioned in these topics. These works show that phrases and knowledge about hyponyms/hypernyms are important for topic representation.
In this paper we describe an approach to integrate large manual lexical resources such as WordNet or EuroVoc into probabilistic topic models, as well as automatically extracted n-grams to improve coherence and informativeness of generated topics. The structure of the paper is as follows. In Section 2 we consider related works. Section 3 describes the proposed approach. Section 4 enumerates automatic quality measures used in experiments. Section 5 presents the results obtained on several text collections according to automatic measures. Section 6 describes the results of manual evaluation of combined topic models for Islam Internet-site thematic analysis.
The authors of the experiment concluded that manual topic labels include a considerable number of phrases; users prefer shorter labels with more general words and tend to incorporate phrases and more generic terminology when using more complex network graph.
|
Do the users tend to incorporate phrases and more generic terminology when using more complex network graphs?
|
Yes.
|
1909.00694
| false
| null |
In this paper, we propose a simple and effective method for learning affective events that only requires a very small seed lexicon and a large raw corpus. As illustrated in Figure FIGREF1, our key idea is that we can exploit discourse relations BIBREF4 to efficiently propagate polarity from seed predicates that directly report one's emotions (e.g., “to be glad” is positive). Suppose that events $x_1$ are $x_2$ are in the discourse relation of Cause (i.e., $x_1$ causes $x_2$). If the seed lexicon suggests $x_2$ is positive, $x_1$ is also likely to be positive because it triggers the positive emotion. The fact that $x_2$ is known to be negative indicates the negative polarity of $x_1$. Similarly, if $x_1$ and $x_2$ are in the discourse relation of Concession (i.e., $x_2$ in spite of $x_1$), the reverse of $x_2$'s polarity can be propagated to $x_1$. Even if $x_2$'s polarity is not known in advance, we can exploit the tendency of $x_1$ and $x_2$ to be of the same polarity (for Cause) or of the reverse polarity (for Concession) although the heuristic is not exempt from counterexamples. We transform this idea into objective functions and train neural network models that predict the polarity of a given event.
The seed lexicon consists of positive and negative predicates. If the predicate of an extracted event is in the seed lexicon and does not involve complex phenomena like negation, we assign the corresponding polarity score ($+1$ for positive events and $-1$ for negative events) to the event. We expect the model to automatically learn complex phenomena through label propagation. Based on the availability of scores and the types of discourse relations, we classify the extracted event pairs into the following three types.
As illustrated in Figure FIGREF1, our key idea is that we can exploit discourse relations BIBREF4 to efficiently propagate polarity from seed predicates that directly report one's emotions (e.g., “to be glad” is positive). Suppose that events $x_1$ are $x_2$ are in the discourse relation of Cause (i.e., $x_1$ causes $x_2$). If the seed lexicon suggests $x_2$ is positive, $x_1$ is also likely to be positive because it triggers the positive emotion. The fact that $x_2$ is known to be negative indicates the negative polarity of $x_1$. Similarly, if $x_1$ and $x_2$ are in the discourse relation of Concession (i.e., $x_2$ in spite of $x_1$), the reverse of $x_2$'s polarity can be propagated to $x_1$. Even if $x_2$'s polarity is not known in advance, we can exploit the tendency of $x_1$ and $x_2$ to be of the same polarity (for Cause) or of the reverse polarity (for Concession) although the heuristic is not exempt from counterexamples. We transform this idea into objective functions and train neural network models that predict the polarity of a given event.
The seed lexicon consists of positive and negative predicates. If the predicate of an extracted event is in the seed lexicon and does not involve complex phenomena like negation, we assign the corresponding polarity score ($+1$ for positive events and $-1$ for negative events) to the event. We expect the model to automatically learn complex phenomena through label propagation.
|
How are relations used to propagate polarity?
|
cause relation: both events in the relation should have the same polarity; concession relation: events should have opposite polarity
|
null | false
| null |
When is the best time to visit San Francisco?
|
The best time to visit San Francisco is right after summer, in September or October. Mark Twain once said the coldest winter he ever spent was a summer in San Francisco. That is because San Francisco has an Indian Summer, when the traditional Summer months are actually quite cold and windy. Starting in September and through October the weather is very moderate and usually sunny.
|
|
null | false
| null |
Why don't soccer players kick with their toe?
|
Soccer players typically strike the ball with their laces or with the inside of their foot. These techniques provide more control over the direction of the ball than kicking with the toe.
|
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.