text stringlengths 0 164k |
|---|
it sigma mathbf WiXX mathbf Wicct1 mathbf bi |
ft sigma mathbf WfXX mathbf Wfcct1 mathbf bf |
ot sigma mathbf WoXX mathbf Wocct1 mathbf bo |
ct ft odot ct1 it odot tanhmathbf WcXX mathbf bc |
ht ot odot tanhct Eq 3 where it ft ot are input forget and output gates respectively The sigmoid function sigma is a soft gate function controlling the amount of information flow W s and b s are model parameters to learn A traditional RNN encoderdecoder approach BIBREF11 first encodes an arbitrary length input s... |
<s>[INST] Attentional Encoder Network for Targeted Sentiment Classification [/INST] Targeted sentiment classification is a finegrained sentiment analysis task which aims at determining the sentiment polarities eg negative neutral or positive of a sentence over opinion targets that explicitly appear in the sentence For ... |
<s>[INST] ThisIsCompetition at SemEval-2019 Task 9: BERT is unstable for out-of-domain samples [/INST] Opinion mining BIBREF0 is a huge field that covers many NLP tasks ranging from sentiment analysis BIBREF1 aspect extraction BIBREF2 and opinion summarization BIBREF3 among others Despite the vast literature on opin... |
<s>[INST] DENS: A Dataset for Multi-class Emotion Analysis [/INST] Humans experience a variety of complex emotions in daily life These emotions are heavily reflected in our language in both spoken and written forms Many recent advances in natural language processing on emotions have focused on product reviews BIBREF0 a... |
<s>[INST] Multitask Learning with CTC and Segmental CRF for Speech Recognition [/INST] Stateoftheart speech recognition accuracy has significantly improved over the past few years since the application of deep neural networks BIBREF0 BIBREF1 Recently it has been shown that with the application of both neural network ... |
<s>[INST] Filling Gender&Number Gaps in Neural Machine Translation with Black-box Context Injection [/INST] A common way for marking information about gender number and case in language is morphology or the structure of a given word in the language However different languages mark such information in different ways fo... |
<s>[INST] Exploring End-to-End Techniques for Low-Resource Speech Recognition [/INST] Although development of the first speech recognition systems began half a century ago there has been a significant increase of the accuracy of ASR systems and number of their applications for the recent ten years even for lowresource ... |
<s>[INST] Tag-based Multi-Span Extraction in Reading Comprehension [/INST] The task of reading comprehension where systems must understand a single passage of text well enough to answer arbitrary questions about it has seen significant progress in the last few years With models reaching human performance on the popular... |
<s>[INST] Transfer Learning Between Related Tasks Using Expected Label Proportions [/INST] Data annotation is a key bottleneck in many data driven algorithms Specifically deep learning models which became a prominent tool in many data driven tasks in recent years require large datasets to work well However many tasks r... |
<s>[INST] The SIGMORPHON 2019 Shared Task: Morphological Analysis in Context and Cross-Lingual Transfer for Inflection [/INST] While producing a sentence humans combine various types of knowledge to produce fluent outputvarious shades of meaning are expressed through word selection and tone while the language is made t... |
<s>[INST] Hierarchical Multi-Task Natural Language Understanding for Cross-domain Conversational AI: HERMIT NLU [/INST] Research in Conversational AI also known as Spoken Dialogue Systems has applications ranging from home devices to robotics and has a growing presence in industry A key problem in realworld Dialogue Sy... |
<s>[INST] Interactive Machine Comprehension with Information Seeking Agents [/INST] Many machine reading comprehension MRC datasets have been released in recent years BIBREF0 BIBREF1 BIBREF2 BIBREF3 BIBREF4 to benchmark a systems ability to understand and reason over natural language Typically these datasets require an... |
<s>[INST] Exploring Hate Speech Detection in Multimodal Publications [/INST] Social Media platforms such as Facebook Twitter or Reddit have empowered individuals voices and facilitated freedom of expression However they have also been a breeding ground for hate speech and other types of online harassment Hate speech is... |
<s>[INST] Self-Taught Convolutional Neural Networks for Short Text Clustering [/INST] Short text clustering is of great importance due to its various applications such as user profiling BIBREF0 and recommendation BIBREF1 for nowadays social media dataset emerged day by day However short text clustering has the data sp... |
<s>[INST] Solving Arithmetic Word Problems Automatically Using Transformer and Unambiguous Representations [/INST] Students are exposed to simple arithmetic word problems starting in elementary school and most become proficient in solving them at a young age Automatic solvers of such problems could potentially help edu... |
<s>[INST] What Do You Mean I'm Funny? Personalizing the Joke Skill of a Voice-Controlled Virtual Assistant [/INST] Voicecontrolled virtual assistants VVA such as Siri and Alexa have experienced an exponential growth in terms of number of users and provided capabilities They are used by millions for a variety of tasks i... |
<s>[INST] A Measure of Similarity in Textual Data Using Spearman's Rank Correlation Coefficient [/INST] Over the past few years the term big data has become an important key point for research into data mining and information retrieval Through the years the quantity of data managed across enterprises has evolved from a... |
<s>[INST] CamemBERT: a Tasty French Language Model [/INST] Pretrained word representations have a long history in Natural Language Processing NLP from nonneural methods BIBREF0 BIBREF1 BIBREF2 to neural word embeddings BIBREF3 BIBREF4 and to contextualised representations BIBREF5 BIBREF6 Approaches shifted more recentl... |
<s>[INST] Vocabulary-based Method for Quantifying Controversy in Social Media [/INST] Controversy is a phenomenom with a high impact at various levels It has been broadly studied from the perspective of different disciplines ranging from the seminal analysis of the conflicts within the members of a karate club BIBREF0 ... |
<s>[INST] Semantic Sentiment Analysis of Twitter Data [/INST] Microblog sentiment analysis Twitter opinion mining Sentiment Analysis This is text analysis aiming to determine the attitude of a speaker or a writer with respect to some topic or the overall contextual polarity of a piece of text Sentiment analysis on Twit... |
<s>[INST] COSTRA 1.0: A Dataset of Complex Sentence Transformations [/INST] Vector representations are becoming truly essential in majority of natural language processing tasks Word embeddings became widely popular with the introduction of word2vec BIBREF0 and GloVe BIBREF1 and their properties have been analyzed in le... |
<s>[INST] Learning to Create Sentence Semantic Relation Graphs for Multi-Document Summarization [/INST] Todays increasing flood of information on the web creates a need for automated multidocument summarization systems that produce high quality summaries However producing summaries in a multidocument setting is difficu... |
<s>[INST] A Deep Neural Architecture for Sentence-level Sentiment Classification in Twitter Social Networking [/INST] Twitter sentiment classification have intensively researched in recent years BIBREF0 BIBREF1 Different approaches were developed for Twitter sentiment classification by using machine learning such as S... |
<s>[INST] Logic Attention Based Neighborhood Aggregation for Inductive Knowledge Graph Embedding [/INST] Knowledge graphs KGs such as Freebase BIBREF0 DBpedia BIBREF1 and YAGO BIBREF2 play a critical role in various NLP tasks including question answering BIBREF3 information retrieval BIBREF4 and personalized recomm... |
<s>[INST] Learning with Noisy Labels for Sentence-level Sentiment Classification [/INST] It is well known that sentiment annotation or labeling is subjective BIBREF0 Annotators often have many disagreements This is especially so for crowdworkers who are not well trained That is why one always feels that there are many ... |
<s>[INST] Keep Calm and Switch On! Preserving Sentiment and Fluency in Semantic Text Exchange [/INST] There has been significant research on style transfer with the goal of changing the style of text while preserving its semantic content The alternative where semantics are adjusted while keeping style intact which we c... |
<s>[INST] CN-CELEB: a challenging Chinese speaker recognition dataset [/INST] Speaker recognition including identification and verification aims to recognize claimed identities of speakers After decades of research performance of speaker recognition systems has been vastly improved and the technique has been deployed t... |
<s>[INST] Conditional BERT Contextual Augmentation [/INST] Deep neural networkbased models are easy to overfit and result in losing their generalization due to limited size of training data In order to address the issue data augmentation methods are often applied to generate more training samples Recent years have witn... |
<s>[INST] Recent Advances in Neural Question Generation [/INST] Question Generation QG concerns the task of automatically generating questions from various inputs such as raw text database or semantic representation BIBREF0 People have the ability to ask rich creative and revealing questions BIBREF1 eg asking Why did... |
vspace1422636pt |
arg max Y sum t1m Pyt vert X A y t Eq 5 BIBREF47 pioneered the first NQG model using an attention Seq2Seq model BIBREF22 which feeds a sentence into an RNNbased encoder and generate a question about the sentence through a decoder The attention mechanism is applied to help decoder pay attention to the most releva... |
<s>[INST] Open Named Entity Modeling from Embedding Distribution [/INST] Named Entity Recognition is a major natural language processing task that recognizes the proper labels such as LOC Location PER Person ORG Organization etc Like words or phrase being a sort of language constituent named entities also benefit from ... |
E O2 Zi K E O1 Xi quad r2 K r1 |
endaligned Eq 17 Given the seed pairs lbrace Xi Zirbrace i1m the initialized center O2 in Equation 3 the center O1 and radius r1 of the hypersphere in source language space we may work out the optimized ratio K the mapping center O2 and radius r2 in target language space by solving the linear equation grou... |
P fracViVm quad R fracViVt quad F1 frac2 P RP R |
endaligned Eq 29 where Vt Vm Vi represent the volumes of the target mapping and intersection hyperspheres Due to the difficulty of calculating the volume of hyperspheres in high dimensions we adopt Monte Carlo methods to simulate the volume BIBREF25 we generate a great quantity of points in the embedding spaces and... |
<s>[INST] Efficient Twitter Sentiment Classification using Subjective Distant Supervision [/INST] A lot of work has been done in the field of Twitter sentiment analysis till date Sentiment analysis has been handled as a Natural Language Processing task at many levels of granularity Most of these techniques use Machine ... |
<s>[INST] Dynamic Memory Networks for Visual and Textual Question Answering [/INST] Neural network based methods have made tremendous progress in image and text classification BIBREF0 BIBREF1 However only recently has progress been made on more complex tasks that require logical reasoning This success is based in par... |
ri sigma leftWrxi Ur hi1 br right |
tildehi tanh leftWxi ri circ U hi1 bhright |
hi uicirc tildehi 1ui circ hi1 Eq 2 where sigma is the sigmoid activation function circ is an elementwise product Wz Wr W in mathbb RnH times nI Uz Ur U in mathbb RnH times nH nH is the hidden size and nI is the input size Question Module This module computes a vector representation q of the question where q i... |
overleftarrowfi GRUbwdfi overleftarrowfi1 |
overleftrightarrowfi overleftarrowfi overrightarrowfi Eq 5 where fi is the input fact at timestep i overrightarrowfi is the hidden state of the forward GRU at timestep i and overleftarrowfi is the hidden state of the backward GRU at timestep i This allows contextual information from both future and past facts ... |
Zti W2 tanh leftW1zti b1 right b2 |
gti fracexp Ztisum k1Mi exp Ztk Eq 10 where overleftrightarrowfi is the ith fact mt1 is the previous episode memory q is the original question circ is the elementwise product cdot is the elementwise absolute value and represents concatenation of the vectors The DMN implemented in BIBREF6 involved a more complex... |
<s>[INST] Low-Level Linguistic Controls for Style Transfer and Content Preservation [/INST] All text has style whether it be formal or informal polite or aggressive colloquial persuasive or even robotic Despite the success of style transfer in image processing BIBREF0 BIBREF1 there has been limited progress in the text... |
<s>[INST] Fusing Visual, Textual and Connectivity Clues for Studying Mental Health [/INST] 0pt00 0pt00 0pt00 095 1Amir Hossein Yazdavar 1Mohammad Saeid Mahdavinejad 2Goonmeet Bajaj 3William Romine 1Amirhassan Monadjemi 1Krishnaprasad Thirunarayan 1Amit Sheth 4Jyotishman Pathak 1Department of Computer Science Enginee... |
<s>[INST] Incorporating Sememes into Chinese Definition Modeling [/INST] Chinese definition modeling is the task of generating a definition in Chinese for a given Chinese word This task can benefit the compilation of dictionaries especially dictionaries for Chinese as a foreign language CFL learners In recent years the... |
mathbf zt fmathbf zt1yt1mathbf x Eq 4 where mathbf zt is the decoders hidden state at time t f is a recurrent nonlinear function such as LSTM and GRU and mathbf x is the embedding of the word being defined Then the probability of Py x can be computed according to the probability chain rule Py x prod t1T Pyty... |
mathbf zt fmathbf zt1yt1mathbf ct Eq 17 where mathbf ct is the context vector from the output of the adaptive attention module at time t mathbf zt is a decoders hidden state at time t To obtain the context vector mathbf ct we first compute the sememe context vector hatmathbf ct and the LM context mathbf ot S... |
etn mathbf whatcTmathbf hn mathbf zt1 Eq 19 Since the decoders hidden states store syntax and semantic information for language modeling we compute the LM context mathbf ot with a gated unit whose input is the definition word yt and the previous hidden state mathbf zt1 mathbf gt sigma mathbf Wg yt1 mathbf zt1 m... |
mathbf ot mathbf gt odot mathrm tanh mathbf zt1 Eq 20 Once the sememe context vector hatmathbf ct and the LM context mathbf ot are ready we can generate the context vector with an adaptive attention layer as mathbf ct beta t mathbf ot 1beta thatmathbf ct Eq 21 where beta t fracmathrm expetomathrm expetoma... |
eto mathbf wcTmathbf otmathbf zt nonumber |
ethatc mathbf wcThatmathbf ctmathbf zt Eq 22 beta t is a scalar in range 01 which controls the relative importance of LM context and sememe context Once we obtain the context vector mathbf ct we can update the decoders hidden state and generate the next word according to Eq and Eq 17 respectively Recent works ... |
leftlbrace beginarrayll |
mathbf x n0 cr mathbf sn n0 |
endarrayright Eq 25 where mathbf x is the vector representation of the given word x and mathbf sn is the vector representation of the n th sememe sn Although the input sequence is not time ordered position n in the sequence carries some useful information First position 0 corresponds to the word to be defined whil... |
etol mathbf wclTmathbf otlmathbf ztl1 nonumber |
ethatcl mathbf wclThatmathbf ctlmathbf ztl1 Eq 31 In this section we will first introduce the construction process of the CDM dataset then the experimental results and analysis To verify our proposed models we construct the CDM dataset for the Chinese definition modeling task cdmEach entry in the dataset is a trip... |
<s>[INST] RobBERT: a Dutch RoBERTa-based Language Model [/INST] The advent of neural networks in natural language processing NLP has significantly improved stateoftheart results within the field While recurrent neural networks RNNs and long shortterm memory networks LSTMs initially dominated the field recent models sta... |
<s>[INST] Natural Language State Representation for Reinforcement Learning [/INST] The world of our experiences must be enormously simplified and generalized before it is possible to make a symbolic inventory of all our experiences of things and relations Edward Sapir Language An Introduction to the Study of Speech 192... |
<s>[INST] Query-oriented text summarization based on hypergraph transversals [/INST] The development of automatic tools for the summarization of large corpora of documents has attracted a widespread interest in recent years With fields of application ranging from medical sciences to finance and legal science these summ... |
V lbrace 1Nsrbrace text and phi iLitext text 1le ile Ns |
E lbrace e1eKrbrace subseteq 2V |
elTltext ie elin textincileftrightarrow iin Tl |
endarray Eq 17 Finally the weights of the hyperedges are computed based on the centrality of the associated theme and its similarity with the query wl1lambda textsimTlDlambda textsimTlq Eq 18 where lambda in 01 is a parameter and D represents the entire corpus textsimTlD denotes the similarity of the set of sent... |
SundersetSsubseteq VtextargminS |
textst undersetiin Sbigcup textinciE |
endarray Eq 21 where textincilbrace eiin erbrace denotes the set of hyperedges incident to node i Figure 2 shows an example of hypergraph and a minimal hypergraph transversal of it starshaped nodes In this case since the nodes and the hyperedges are unweighted the minimal transversal is not unique The problem of f... |
SundersetSsubseteq Vtextargminundersetiin Ssum phi i |
textst undersetein textincSsum wege gamma W |
endarray Eq 24 in which textincSundersetiin Sbigcup textinci and Wsum ewe The extraction of a minimal soft hypergraph transversal of the sentence hypergraph produces a summary of minimal length achieving a target coverage expressed by parameter gamma in 01 As mentioned in section Problem statement and system overv... |
SundersetSsubseteq Vtextargmaxundersetein textincSsum we |
textst undersetiin Ssum phi ile L |
endarray Eq 26 We refer to the function undersetein textincSsum we as the hyperedge coverage of set S We observe that both weighted transversals defined above include the notion of joint coverage of the hyperedges by the selected nodes As a result and from the definition of hyperedge weights equation 18 the result... |
fScup lbrace srbrace fSfTcup lbrace srbrace fT |
leftundersetein textincScup lbrace srbrace sum weundersetein textincSsum werightleftundersetein textincTcup lbrace srbrace sum weundersetein textincTsum weright |
left undersetein textinclbrace srbrace setminus textincSsum werightleft undersetein textinclbrace srbrace setminus textincTsum weright |
undersetein textincTcap textinclbrace srbrace setminus textincSsum wege 0 |
endarray Eq 33 where textincRlbrace eecap Sne emptyset rbrace for Rsubseteq V The last equality follows from textincSsubseteq textincT and textinclbrace srbrace setminus textincTsubseteq textinclbrace srbrace setminus textincS Various classes of NPhard problems involving a submodular and nondecreasing function ca... |
<s>[INST] Text-based inference of moral sentiment change [/INST] Peoples moral sentimentour feelings toward right or wrongcan change over time For instance the publics views toward slavery have shifted substantially over the past centuries BIBREF0 How societys moral views evolve has been a longstanding issue and a cons... |
<s>[INST] Bringing Stories Alive: Generating Interactive Fiction Worlds [/INST] Interactive fictionsalso called textadventure games or textbased gamesare games in which a player interacts with a virtual world purely through textual natural languagereceiving descriptions of what they see and writing out how they want to... |
<s>[INST] Generating Classical Chinese Poems from Vernacular Chinese [/INST] During thousands of years millions of classical Chinese poems have been written They contain ancient poets emotions such as their appreciation for nature desiring for freedom and concerns for their countries Among various types of classical po... |
<s>[INST] Entity-Consistent End-to-end Task-Oriented Dialogue System with KB Retriever [/INST] Taskoriented dialogue system which helps users to achieve specific goals with natural language is attracting more and more research attention With the success of the sequencetosequence Seq2Seq models in text generation BIBREF... |
<s>[INST] From FiLM to Video: Multi-turn Question Answering with Multi-modal Context [/INST] Deep neural networks have been successfully applied to several computer vision tasks such as image classification BIBREF0 object detection BIBREF1 video action classification BIBREF2 etc They have also been successfully appl... |
<s>[INST] Civique: Using Social Media to Detect Urban Emergencies [/INST] With the surge in the use of social media microblogging sites like Twitter Facebook and Foursquare have become household words Growing ubiquity of mobile phones in highly populated developing nations has spurred an exponential rise in social medi... |
<s>[INST] Can neural networks understand monotonicity reasoning? [/INST] Natural language inference NLI also known as recognizing textual entailment RTE has been proposed as a benchmark task for natural language understanding Given a premise P and a hypothesis H the task is to determine whether the premise semanticall... |
<s>[INST] Enriching Existing Conversational Emotion Datasets with Dialogue Acts using Neural Annotators. [/INST] With the growing demand for humancomputerrobot interaction systems detecting the emotional state of the user can heavily benefit a conversational agent to respond at an appropriate emotional level Emotion re... |
<s>[INST] Synchronising audio and ultrasound by learning cross-modal embeddings [/INST] Ultrasound tongue imaging UTI is a noninvasive way of observing the vocal tract during speech production BIBREF0 Instrumental speech therapy relies on capturing ultrasound videos of the patients tongue simultaneously with their spe... |
<s>[INST] Basic tasks of sentiment analysis [/INST] School of Computer Science and Engineering Nanyang Technological University Singapore Sentiment Analysis Subjectivity Detection Deep Learning Aspect Extraction Polarity Distribution Convolutional Neural Network Aspect Feature related to an opinion target Convolution ... |
<s>[INST] Generalisation in Named Entity Recognition: A Quantitative Analysis [/INST] Named entity recognition and classification NERC short NER the task of recognising and assigning a class to mentions of proper names named entities NEs in text has attracted many years of research BIBREF0 BIBREF1 analyses BIBREF2 s... |
<s>[INST] wav2vec: Unsupervised Pre-training for Speech Recognition [/INST] Current state of the art models for speech recognition require large amounts of transcribed audio data to attain good performance BIBREF1 Recently pretraining of neural networks has emerged as an effective technique for settings where labeled ... |
<s>[INST] Cross-lingual, Character-Level Neural Morphological Tagging [/INST] Stateoftheart morphological taggers require thousands of annotated sentences to train For the majority of the worlds languages however sufficient largescale annotation is not available and obtaining it would often be infeasible Accordingly an... |
textLSTMvi1Nright Eq 17 where each vi in mathbb Rn is itself a word embedding Note that the function textLSTM returns the last final hidden state vector of the network This architecture is the context bidirectional recurrent neural network of planksogaardgoldberg2016P162 Finally we derive each word embedding vector... |
ciMirangle right right |
left textLSTM leftlangle ciMi ldots ci1rangle right right nonumber Eq 18 In other words we run a bidirectional LSTM over the character stream This bidirectional LSTM is the sequence bidirectional recurrent neural network of planksogaardgoldberg2016P162 Note a concatenation of the sequence of character symbols la... |
Subsets and Splits
No community queries yet
The top public SQL queries from the community will appear here once available.