Unnamed: 0.1
int64
0
41k
Unnamed: 0
int64
0
41k
author
stringlengths
9
1.39k
id
stringlengths
11
18
summary
stringlengths
25
3.66k
title
stringlengths
4
258
year
int64
1.99k
2.02k
arxiv_url
stringlengths
32
39
info
stringlengths
523
3.18k
embeddings
stringlengths
16.9k
17.1k
1,600
1,600
['Nicholas H. Kirk']
1410.8326v1
Everyday activities performed by artificial assistants can potentially be executed naively and dangerously given their lack of common sense knowledge. This paper presents conceptual work towards obtaining prior knowledge on the usual modality (passive or active) of any given entity, and their affordance estimates, by e...
Towards Learning Object Affordance Priors from Technical Texts
2,014
http://arxiv.org/pdf/1410.8326v1
Title Towards Learning Object Affordance Priors Technical Texts Summary Everyday activity performed artificial assistant potentially executed naively dangerously given lack common sense knowledge paper present conceptual work towards obtaining prior knowledge usual modality passive active given entity affordance estima...
[0.03057178109884262, 0.039893485605716705, -0.008277135901153088, 0.02691507153213024, -0.01073034293949604, 0.01521222572773695, 0.024952441453933716, 0.004894418176263571, -0.008545022457838058, -0.058323513716459274, 0.0228132177144289, 0.03413478657603264, 0.0005273465067148209, 0.040852636098861694, -0.0124802589...
1,601
1,601
['Jiwei Li', 'Alan Ritter', 'Dan Jurafsky']
1411.2679v1
We propose a framework for inferring the latent attitudes or preferences of users by performing probabilistic first-order logical reasoning over the social network graph. Our method answers questions about Twitter users like {\em Does this user like sushi?} or {\em Is this user a New York Knicks fan?} by building a pro...
Inferring User Preferences by Probabilistic Logical Reasoning over Social Networks
2,014
http://arxiv.org/pdf/1411.2679v1
Title Inferring User Preferences Probabilistic Logical Reasoning Social Networks Summary propose framework inferring latent attitude preference user performing probabilistic firstorder logical reasoning social network graph method answer question Twitter user like em user like sushi em user New York Knicks fan building...
[0.05918864533305168, 0.07159703224897385, -0.003193802898749709, 0.006791557185351849, -0.01331583596765995, 0.01091077458113432, 0.0032681815791875124, -0.005324776750057936, 0.017179055139422417, -0.03386303782463074, 0.028839662671089172, 0.030044453218579292, -0.0065878466702997684, 0.049059830605983734, -0.006814...
1,602
1,602
['Piotr Mirowski', 'Andreas Vlachos']
1507.01193v1
Recent work on language modelling has shifted focus from count-based models to neural models. In these works, the words in each sentence are always considered in a left-to-right order. In this paper we show how we can improve the performance of the recurrent neural network (RNN) language model by incorporating the synt...
Dependency Recurrent Neural Language Models for Sentence Completion
2,015
http://arxiv.org/pdf/1507.01193v1
Title Dependency Recurrent Neural Language Models Sentence Completion Summary Recent work language modelling shifted focus countbased model neural model work word sentence always considered lefttoright order paper show improve performance recurrent neural network RNN language model incorporating syntactic dependency se...
[0.06822627037763596, 0.05796389281749725, -0.00980827771127224, 0.051595523953437805, -0.056541286408901215, -0.030912617221474648, -0.017460057511925697, -0.010839520022273064, 0.015038248151540756, -0.019280023872852325, 0.02414649724960327, -0.025141064077615738, 0.03717095032334328, 0.031153259798884392, 0.0038425...
1,603
1,603
['Mingbo Ma', 'Liang Huang', 'Bing Xiang', 'Bowen Zhou']
1507.01839v2
In sentence modeling and classification, convolutional neural network approaches have recently achieved state-of-the-art results, but all such efforts process word vectors sequentially and neglect long-distance dependencies. To exploit both deep learning and linguistic structures, we propose a tree-based convolutional ...
Dependency-based Convolutional Neural Networks for Sentence Embedding
2,015
http://arxiv.org/pdf/1507.01839v2
Title Dependencybased Convolutional Neural Networks Sentence Embedding Summary sentence modeling classification convolutional neural network approach recently achieved stateoftheart result effort process word vector sequentially neglect longdistance dependency exploit deep learning linguistic structure propose treebase...
[0.06396469473838806, 0.0691571980714798, 0.0019338791025802493, 0.10003326833248138, -0.053652264177799225, 0.007408928591758013, -0.02438923716545105, -0.01956883631646633, 0.008519395254552364, -0.05386137589812279, 0.021681323647499084, 0.001110645243898034, -0.020862944424152374, 0.013806281611323357, -0.024739935...
1,604
1,604
['Yang Yu', 'Wei Zhang', 'Chung-Wei Hang', 'Bing Xiang', 'Bowen Zhou']
1510.07526v3
In this paper we explore deep learning models with memory component or attention mechanism for question answering task. We combine and compare three models, Neural Machine Translation, Neural Turing Machine, and Memory Networks for a simulated QA data set. This paper is the first one that uses Neural Machine Translatio...
Empirical Study on Deep Learning Models for Question Answering
2,015
http://arxiv.org/pdf/1510.07526v3
Title Empirical Study Deep Learning Models Question Answering Summary paper explore deep learning model memory component attention mechanism question answering task combine compare three model Neural Machine Translation Neural Turing Machine Memory Networks simulated QA data set paper first one us Neural Machine Transl...
[0.04352615401148796, -0.017084674909710884, -0.02108597569167614, 0.019770175218582153, -0.016032980754971504, 0.022740868851542473, 0.008079138584434986, -0.020140843465924263, -0.010329513810575008, -0.0344117097556591, -0.010545307770371437, -0.016429748386144638, -0.01855245605111122, 0.010898835025727749, 0.02351...
1,605
1,605
['Ji He', 'Jianshu Chen', 'Xiaodong He', 'Jianfeng Gao', 'Lihong Li', 'Li Deng', 'Mari Ostendorf']
1511.04636v5
This paper introduces a novel architecture for reinforcement learning with deep neural networks designed to handle state and action spaces characterized by natural language, as found in text-based games. Termed a deep reinforcement relevance network (DRRN), the architecture represents action and state spaces with separ...
Deep Reinforcement Learning with a Natural Language Action Space
2,015
http://arxiv.org/pdf/1511.04636v5
Title Deep Reinforcement Learning Natural Language Action Space Summary paper introduces novel architecture reinforcement learning deep neural network designed handle state action space characterized natural language found textbased game Termed deep reinforcement relevance network DRRN architecture represents action st...
[0.07113111019134521, -0.017569445073604584, -0.03022187389433384, 0.04427012428641319, -0.031326647847890854, 0.004467928782105446, -0.03132466599345207, 0.0017369576962664723, 0.016057677567005157, -0.029971245676279068, -0.030482100322842598, -0.028076739981770515, -0.03135330602526665, 0.07112657278776169, 0.004616...
1,606
1,606
['Volker Tresp', 'Cristóbal Esteban', 'Yinchong Yang', 'Stephan Baier', 'Denis Krompaß']
1511.07972v9
Embedding learning, a.k.a. representation learning, has been shown to be able to model large-scale semantic knowledge graphs. A key concept is a mapping of the knowledge graph to a tensor representation whose entries are predicted by models using latent representations of generalized entities. Latent variable models ar...
Learning with Memory Embeddings
2,015
http://arxiv.org/pdf/1511.07972v9
Title Learning Memory Embeddings Summary Embedding learning aka representation learning shown able model largescale semantic knowledge graph key concept mapping knowledge graph tensor representation whose entry predicted model using latent representation generalized entity Latent variable model well suited deal high di...
[0.02625063806772232, -0.04695550724864006, -0.0364769846200943, 0.05098911374807358, 0.01442122645676136, 0.029984649270772934, -0.013913542032241821, 0.0016957212937995791, 0.05053820461034775, -0.07250671088695526, 0.02826927974820137, 0.014758506789803505, -0.0020753885619342327, 0.0006650813738815486, 0.0314340330...
1,607
1,607
['Jiaxin Shi', 'Jun Zhu']
1512.01173v1
We present a new perspective on neural knowledge base (KB) embeddings, from which we build a framework that can model symbolic knowledge in the KB together with its learning process. We show that this framework well regularizes previous neural KB embedding model for superior performance in reasoning tasks, while having...
Building Memory with Concept Learning Capabilities from Large-scale Knowledge Base
2,015
http://arxiv.org/pdf/1512.01173v1
Title Building Memory Concept Learning Capabilities Largescale Knowledge Base Summary present new perspective neural knowledge base KB embeddings build framework model symbolic knowledge KB together learning process show framework well regularizes previous neural KB embedding model superior performance reasoning task c...
[0.05125345289707184, -0.01162670087069273, -0.018074050545692444, 0.047930553555488586, 0.016361357644200325, 0.013930734246969223, -0.021893782541155815, 0.0031508977990597486, -0.019199438393115997, -0.04718167707324028, -0.017136089503765106, 0.006309432443231344, -0.01319514773786068, 0.05573263391852379, 0.039557...
1,608
1,608
['Kamil Rocki']
1512.01926v1
There exists a theory of a single general-purpose learning algorithm which could explain the principles its operation. It assumes the initial rough architecture, a small library of simple innate circuits which are prewired at birth. and proposes that all significant mental algorithms are learned. Given current understa...
Thinking Required
2,015
http://arxiv.org/pdf/1512.01926v1
Title Thinking Required Summary exists theory single generalpurpose learning algorithm could explain principle operation assumes initial rough architecture small library simple innate circuit prewired birth proposes significant mental algorithm learned Given current understanding observation paper review list ingredien...
[-0.03977697715163231, 0.012577125802636147, -0.039064135402441025, 0.0010032934369519353, -0.01318944338709116, -0.008451482281088829, 0.007296303287148476, -0.0005464830319397151, 0.005177484359592199, 0.014293735846877098, -0.0032003852538764477, 0.031205350533127785, -0.0063491202890872955, 0.024950893595814705, 0....
1,609
1,609
['Pierre-Yves Oudeyer']
1601.00816v1
This article discusses open scientific challenges for understanding development and evolution of speech forms, as a commentary to Moulin-Frier et al. (Moulin-Frier et al., 2015). Based on the analysis of mathematical models of the origins of speech forms, with a focus on their assumptions , we study the fundamental que...
Open challenges in understanding development and evolution of speech forms: The roles of embodied self-organization, motivation and active exploration
2,016
http://arxiv.org/pdf/1601.00816v1
Title Open challenge understanding development evolution speech form role embodied selforganization motivation active exploration Summary article discus open scientific challenge understanding development evolution speech form commentary MoulinFrier et al MoulinFrier et al 2015 Based analysis mathematical model origin ...
[0.026993047446012497, 0.03808136656880379, -0.05852857232093811, -0.030663585290312767, 0.034296780824661255, 0.012988943606615067, -0.00105187704320997, -0.03494023531675339, -0.05445825308561325, -0.03593285381793976, -0.01708143949508667, -0.03978940472006798, 0.028907783329486847, -0.02333667501807213, 0.059513110...
1,610
1,610
['Franck Dernoncourt', 'Ji Young Lee', 'Trung H. Bui', 'Hung H. Bui']
1605.02129v1
The Dialog State Tracking Challenge 4 (DSTC 4) proposes several pilot tasks. In this paper, we focus on the spoken language understanding pilot task, which consists of tagging a given utterance with speech acts and semantic slots. We compare different classifiers: the best system obtains 0.52 and 0.67 F1-scores on the ...
Adobe-MIT submission to the DSTC 4 Spoken Language Understanding pilot task
2,016
http://arxiv.org/pdf/1605.02129v1
Title AdobeMIT submission DSTC 4 Spoken Language Understanding pilot task Summary Dialog State Tracking Challenge 4 DSTC 4 proposes several pilot task paper focus spoken language understanding pilot task consists tagging given utterance speech act semantic slot compare different classifier best system obtains 052 067 F...
[0.003910907544195652, 0.012982658110558987, -0.008425083942711353, 0.052335843443870544, -0.021064141765236855, -0.016579290851950645, -0.002834977814927697, 0.018141139298677444, 0.022610243409872055, -0.0681898295879364, -0.009964067488908768, -0.027133271098136902, 0.015267334878444672, 0.1036924421787262, -0.04808...
1,611
1,611
['Tiancheng Zhao', 'Maxine Eskenazi']
1606.02560v2
This paper presents an end-to-end framework for task-oriented dialog systems using a variant of Deep Recurrent Q-Networks (DRQN). The model is able to interface with a relational database and jointly learn policies for both language understanding and dialog strategy. Moreover, we propose a hybrid algorithm that combine...
Towards End-to-End Learning for Dialog State Tracking and Management using Deep Reinforcement Learning
2,016
http://arxiv.org/pdf/1606.02560v2
Title Towards EndtoEnd Learning Dialog State Tracking Management using Deep Reinforcement Learning Summary paper present endtoend framework taskoriented dialog system using variant Deep Recurrent QNetworks DRQN model able interface relational database jointly learn policy language understanding dialog strategy Moreover...
[0.030138282105326653, 0.040509410202503204, -0.0012052054516971111, 0.013314341194927692, 0.0018608628306537867, 0.009577442891895771, -0.004967265762388706, -0.03496478125452995, -0.007143090944737196, -0.05078567937016487, -0.016546200960874557, -0.055566947907209396, -0.04047919437289238, 0.10333795845508575, -0.01...
1,612
1,612
['Ji He', 'Mari Ostendorf', 'Xiaodong He', 'Jianshu Chen', 'Jianfeng Gao', 'Lihong Li', 'Li Deng']
1606.03667v4
We introduce an online popularity prediction and tracking task as a benchmark task for reinforcement learning with a combinatorial, natural language action space. A specified number of discussion threads predicted to be popular are recommended, chosen from a fixed window of recent comments to track. Novel deep reinforc...
Deep Reinforcement Learning with a Combinatorial Action Space for Predicting Popular Reddit Threads
2,016
http://arxiv.org/pdf/1606.03667v4
Title Deep Reinforcement Learning Combinatorial Action Space Predicting Popular Reddit Threads Summary introduce online popularity prediction tracking task benchmark task reinforcement learning combinatorial natural language action space specified number discussion thread predicted popular recommended chosen fixed wind...
[0.06483180820941925, 0.018846990540623665, -0.02650412544608116, -0.011286293156445026, 0.013935725204646587, -0.014653990045189857, 0.00936222542077303, -0.0008701990591362119, 0.009035184048116207, -0.0150891849771142, -0.011651430279016495, -0.06970905512571335, -0.05350176617503166, 0.09029275923967361, 0.01984805...
1,613
1,613
['Nikola Mrkšić', 'Diarmuid Ó Séaghdha', 'Tsung-Hsien Wen', 'Blaise Thomson', 'Steve Young']
1606.03777v2
One of the core components of modern spoken dialogue systems is the belief tracker, which estimates the user's goal at every step of the dialogue. However, most current approaches have difficulty scaling to larger, more complex dialogue domains. This is due to their dependency on either: a) Spoken Language Understandin...
Neural Belief Tracker: Data-Driven Dialogue State Tracking
2,016
http://arxiv.org/pdf/1606.03777v2
Title Neural Belief Tracker DataDriven Dialogue State Tracking Summary One core component modern spoken dialogue system belief tracker estimate user goal every step dialogue However current approach difficulty scaling larger complex dialogue domain due dependency either Spoken Language Understanding model require large...
[0.020744726061820984, 0.12114806473255157, 0.013274140655994415, 0.060545261949300766, -0.006217618472874165, -0.03607563301920891, -0.019411301240324974, -0.03192145377397537, 0.012556028552353382, -0.027745507657527924, -0.019152013584971428, -0.0861222967505455, -0.02229471132159233, 0.08961950242519379, -0.0161558...
1,614
1,614
['Denis Paperno', 'Germán Kruszewski', 'Angeliki Lazaridou', 'Quan Ngoc Pham', 'Raffaella Bernardi', 'Sandro Pezzelle', 'Marco Baroni', 'Gemma Boleda', 'Raquel Fernández']
1606.06031v1
We introduce LAMBADA, a dataset to evaluate the capabilities of computational models for text understanding by means of a word prediction task. LAMBADA is a collection of narrative passages sharing the characteristic that human subjects are able to guess their last word if they are exposed to the whole passage, but not...
The LAMBADA dataset: Word prediction requiring a broad discourse context
2,016
http://arxiv.org/pdf/1606.06031v1
Title LAMBADA dataset Word prediction requiring broad discourse context Summary introduce LAMBADA dataset evaluate capability computational model text understanding mean word prediction task LAMBADA collection narrative passage sharing characteristic human subject able guess last word exposed whole passage see last sen...
[0.052092913538217545, 0.04661012440919876, -0.012627486139535904, 0.0306735597550869, -0.03002794273197651, -0.018985621631145477, -0.023434750735759735, 0.018596800044178963, -0.033782001584768295, -0.041952911764383316, -0.01851692982017994, -0.015963831916451454, 0.04827887937426567, 0.04258812218904495, 0.01828742...
1,615
1,615
['Fereshte Khani', 'Martin Rinard', 'Percy Liang']
1606.06368v2
Can we train a system that, on any new input, either says "don't know" or makes a prediction that is guaranteed to be correct? We answer the question in the affirmative provided our model family is well-specified. Specifically, we introduce the unanimity principle: only predict when all models consistent with the train...
Unanimous Prediction for 100% Precision with Application to Learning Semantic Mappings
2,016
http://arxiv.org/pdf/1606.06368v2
Title Unanimous Prediction 100 Precision Application Learning Semantic Mappings Summary train system new input either say dont know make prediction guaranteed correct answer question affirmative provided model family wellspecified Specifically introduce unanimity principle predict model consistent training data predict...
[0.05200977995991707, 0.06482750922441483, -0.005546786822378635, 0.032720789313316345, -0.04599402844905853, -0.013249627314507961, 0.011668547987937927, -0.01837318204343319, 0.0205446295440197, -0.05972610041499138, 0.03167935833334923, 0.037686120718717575, 0.02520410344004631, 0.027874812483787537, -0.005419120658...
1,616
1,616
['Thomas Demeester', 'Tim Rocktäschel', 'Sebastian Riedel']
1606.08359v2
Methods based on representation learning currently hold the state-of-the-art in many natural language processing and knowledge base inference tasks. Yet, a major challenge is how to efficiently incorporate commonsense knowledge into such models. A recent approach regularizes relation and entity representations by propo...
Lifted Rule Injection for Relation Embeddings
2,016
http://arxiv.org/pdf/1606.08359v2
Title Lifted Rule Injection Relation Embeddings Summary Methods based representation learning currently hold stateoftheart many natural language processing knowledge base inference task Yet major challenge efficiently incorporate commonsense knowledge model recent approach regularizes relation entity representation pro...
[0.06628160923719406, 0.0280791986733675, 0.008972791023552418, 0.06747307628393173, -0.010398068465292454, 0.003377973334863782, -0.037802670150995255, 0.006237308494746685, 0.06807239353656769, -0.052993860095739365, 0.030892468988895416, 0.051224205642938614, -0.002197754103690386, 0.015358499251306057, 0.0031548654...
1,617
1,617
['Dileep Viswanathan', 'Ameet Soni', 'Jude Shavlik', 'Sriraam Natarajan']
1607.00424v1
We consider the task of KBP slot filling -- extracting relation information from newswire documents for knowledge base construction. We present our pipeline, which employs Relational Dependency Networks (RDNs) to learn linguistic patterns for relation extraction. Additionally, we demonstrate how several components such...
Learning Relational Dependency Networks for Relation Extraction
2,016
http://arxiv.org/pdf/1607.00424v1
Title Learning Relational Dependency Networks Relation Extraction Summary consider task KBP slot filling extracting relation information newswire document knowledge base construction present pipeline employ Relational Dependency Networks RDNs learn linguistic pattern relation extraction Additionally demonstrate several...
[0.07116428762674332, 0.03815937042236328, 0.02583649754524231, 0.08225893974304199, -0.043539293110370636, 0.02634509839117527, -0.04152751341462135, 0.032483018934726715, -0.004982577171176672, -0.028006240725517273, 0.006192058324813843, 0.047861747443675995, -0.03313792124390602, 0.013047580607235432, -0.0251913219...
1,618
1,618
['Emmanuel Dupoux']
1607.08723v4
During their first years of life, infants learn the language(s) of their environment at an amazing speed despite large cross cultural variations in amount and complexity of the available language input. Understanding this simple fact still escapes current cognitive and linguistic theories. Recently, spectacular progres...
Cognitive Science in the era of Artificial Intelligence: A roadmap for reverse-engineering the infant language-learner
2,016
http://arxiv.org/pdf/1607.08723v4
Title Cognitive Science era Artificial Intelligence roadmap reverseengineering infant languagelearner Summary first year life infant learn language environment amazing speed despite large cross cultural variation amount complexity available language input Understanding simple fact still escape current cognitive linguis...
[0.024136599153280258, 0.0513371117413044, -0.06443532556295395, 0.0051691653206944466, -0.030560849234461784, 0.03024349734187126, 0.008981394581496716, 0.015704382210969925, 0.014916922897100449, -0.03643222153186798, 0.014413564465939999, 0.002790896687656641, 0.05124704912304878, 0.05933316424489021, 0.006493132095...
1,619
1,619
['Christophe Van Gysel', 'Maarten de Rijke', 'Marcel Worring']
1608.06651v2
We introduce an unsupervised discriminative model for the task of retrieving experts in online document collections. We exclusively employ textual evidence and avoid explicit feature engineering by learning distributed word representations in an unsupervised way. We compare our model to state-of-the-art unsupervised st...
Unsupervised, Efficient and Semantic Expertise Retrieval
2,016
http://arxiv.org/pdf/1608.06651v2
Title Unsupervised Efficient Semantic Expertise Retrieval Summary introduce unsupervised discriminative model task retrieving expert online document collection exclusively employ textual evidence avoid explicit feature engineering learning distributed word representation unsupervised way compare model stateoftheart uns...
[0.08478772640228271, 0.019152848049998283, 0.008124180138111115, 0.012896285392343998, -0.009221714921295643, -0.03827350214123726, 0.05610213801264763, -0.0038565488066524267, 0.004307198338210583, -0.05575566366314888, -0.034623388200998306, 0.06570344418287277, 0.0018697453197091818, 0.004858138505369425, -0.006253...
1,620
1,620
['Aylin Caliskan', 'Joanna J. Bryson', 'Arvind Narayanan']
1608.07187v4
Artificial intelligence and machine learning are in a period of astounding growth. However, there are concerns that these technologies may be used, either with or without intention, to perpetuate the prejudice and unfairness that unfortunately characterizes many human institutions. Here we show for the first time that ...
Semantics derived automatically from language corpora contain human-like biases
2,016
http://arxiv.org/pdf/1608.07187v4
Title Semantics derived automatically language corpus contain humanlike bias Summary Artificial intelligence machine learning period astounding growth However concern technology may used either without intention perpetuate prejudice unfairness unfortunately characterizes many human institution show first time humanlike...
[0.05592048540711403, 0.06652335822582245, -0.026119688525795937, 0.014433661475777626, -0.0220699030905962, 0.032133664935827255, 0.0476866289973259, -0.0038980757817626, 0.02785390615463257, -0.0786200761795044, 0.01741110160946846, 0.06589112430810928, 0.02537771314382553, -0.004609909374266863, 0.018694257363677025...
1,621
1,621
['Ronan Collobert', 'Christian Puhrsch', 'Gabriel Synnaeve']
1609.03193v2
This paper presents a simple end-to-end model for speech recognition, combining a convolutional network based acoustic model and a graph decoding. It is trained to output letters, with transcribed speech, without the need for force alignment of phonemes. We introduce an automatic segmentation criterion for training fro...
Wav2Letter: an End-to-End ConvNet-based Speech Recognition System
2,016
http://arxiv.org/pdf/1609.03193v2
Title Wav2Letter EndtoEnd ConvNetbased Speech Recognition System Summary paper present simple endtoend model speech recognition combining convolutional network based acoustic model graph decoding trained output letter transcribed speech without need force alignment phoneme introduce automatic segmentation criterion tra...
[-0.014217793010175228, 0.021440984681248665, 0.06417328864336014, 0.07334186881780624, 0.00830492377281189, -0.018732920289039612, -0.009552118368446827, -0.0013938569463789463, -0.02449853904545307, -0.007574591785669327, -0.03466521203517914, 0.006704075261950493, 0.03574751317501068, 0.061495110392570496, 0.0306843...
1,622
1,622
['Lantian Li', 'Yixiang Chen', 'Dong Wang', 'Chenghui Zhao']
1609.08441v2
PLDA is a popular normalization approach for the i-vector model, and it has delivered state-of-the-art performance in speaker verification. However, PLDA training requires a large amount of labelled development data, which is highly expensive in most cases. We present a cheap PLDA training approach, which assumes that ...
Weakly Supervised PLDA Training
2,016
http://arxiv.org/pdf/1609.08441v2
Title Weakly Supervised PLDA Training Summary PLDA popular normalization approach ivector model delivered stateoftheart performance speaker verification However PLDA training requires large amount labelled development data highly expensive case present cheap PLDA training approach assumes speaker session easily separat...
[-0.00714223925024271, -0.0029181567952036858, -0.008761437609791756, 0.03839460015296936, 0.02385341189801693, -0.008656748570501804, 0.03776707500219345, -0.026435233652591705, -0.010988041758537292, 0.03382178023457527, -0.06113220378756523, -0.01789318397641182, 0.041397202759981155, 0.01261075958609581, 0.04731387...
1,623
1,623
['Kaixiang Mo', 'Shuangyin Li', 'Yu Zhang', 'Jiajun Li', 'Qiang Yang']
1610.02891v3
It is difficult to train a personalized task-oriented dialogue system because the data collected from each individual is often insufficient. Personalized dialogue systems trained on a small dataset can overfit and make it difficult to adapt to different user needs. One way to solve this problem is to consider a collect...
Personalizing a Dialogue System with Transfer Reinforcement Learning
2,016
http://arxiv.org/pdf/1610.02891v3
Title Personalizing Dialogue System Transfer Reinforcement Learning Summary difficult train personalized taskoriented dialogue system data collected individual often insufficient Personalized dialogue system trained small dataset overfit make difficult adapt different user need One way solve problem consider collection...
[0.062248773872852325, 0.028949668630957603, -0.020200561732053757, 0.045696090906858444, -0.01140110194683075, 0.000624875829089433, 0.018826980143785477, -0.017685048282146454, -0.021665774285793304, -0.035337094217538834, -0.0774838998913765, 0.03672889992594719, -0.05344853550195694, 0.05756896361708641, -0.0225252...
1,624
1,624
['Chen Liang', 'Jonathan Berant', 'Quoc Le', 'Kenneth D. Forbus', 'Ni Lao']
1611.00020v4
Harnessing the statistical power of neural networks to perform language understanding and symbolic reasoning is difficult, when it requires executing efficient discrete operations against a large knowledge-base. In this work, we introduce a Neural Symbolic Machine, which contains (a) a neural "programmer", i.e., a sequ...
Neural Symbolic Machines: Learning Semantic Parsers on Freebase with Weak Supervision
2,016
http://arxiv.org/pdf/1611.00020v4
Title Neural Symbolic Machines Learning Semantic Parsers Freebase Weak Supervision Summary Harnessing statistical power neural network perform language understanding symbolic reasoning difficult requires executing efficient discrete operation large knowledgebase work introduce Neural Symbolic Machine contains neural pr...
[0.0433107428252697, 0.03959548473358154, -0.02033154107630253, 0.02895292453467846, -0.020251639187335968, 0.031929850578308105, 0.0007463189540430903, -0.004354465752840042, 0.003726033726707101, -0.029881155118346214, 0.02237592823803425, 0.04127662628889084, -0.025979841127991676, 0.05927430838346481, 0.00994130782...
1,625
1,625
['Lajanugen Logeswaran', 'Honglak Lee', 'Dragomir Radev']
1611.02654v2
Modeling the structure of coherent texts is a key NLP problem. The task of coherently organizing a given set of sentences has been commonly used to build and evaluate models that understand such structure. We propose an end-to-end unsupervised deep learning approach based on the set-to-sequence framework to address thi...
Sentence Ordering and Coherence Modeling using Recurrent Neural Networks
2,016
http://arxiv.org/pdf/1611.02654v2
Title Sentence Ordering Coherence Modeling using Recurrent Neural Networks Summary Modeling structure coherent text key NLP problem task coherently organizing given set sentence commonly used build evaluate model understand structure propose endtoend unsupervised deep learning approach based settosequence framework add...
[0.021336128935217857, 0.031323131173849106, 0.007901083678007126, 0.07674999535083771, -0.033915650099515915, -0.003431654768064618, -0.005772426258772612, -0.046165548264980316, -0.01152301486581564, -0.016119273379445076, 0.01953808404505253, 0.003023745259270072, 0.019641758874058723, 0.02435370720922947, -0.032338...
1,626
1,626
['Wei-Fan Chen', 'Lun-Wei Ku']
1611.03599v1
Most neural network models for document classification on social media focus on text infor-mation to the neglect of other information on these platforms. In this paper, we classify post stance on social media channels and develop UTCNN, a neural network model that incorporates user tastes, topic tastes, and user commen...
UTCNN: a Deep Learning Model of Stance Classificationon on Social Media Text
2,016
http://arxiv.org/pdf/1611.03599v1
Title UTCNN Deep Learning Model Stance Classificationon Social Media Text Summary neural network model document classification social medium focus text information neglect information platform paper classify post stance social medium channel develop UTCNN neural network model incorporates user taste topic taste user co...
[0.044409703463315964, 0.0679924413561821, 0.023902202025055885, 0.025464629754424095, -0.034755777567625046, 0.029246626421809196, 0.038479797542095184, -0.01647786982357502, -0.010141825303435326, -0.07947465777397156, 0.010010934434831142, -0.03721160441637039, -0.02992984466254711, 0.02186630666255951, -0.006848441...
1,627
1,627
['Yelong Shen', 'Po-Sen Huang', 'Ming-Wei Chang', 'Jianfeng Gao']
1611.04642v4
Recent studies on knowledge base completion, the task of recovering missing facts based on observed facts, demonstrate the importance of learning embeddings from multi-step relations. Due to the size of knowledge bases, previous works manually design relation paths of observed triplets in symbolic space (e.g. random wa...
Traversing Knowledge Graph in Vector Space without Symbolic Space Guidance
2,016
http://arxiv.org/pdf/1611.04642v4
Title Traversing Knowledge Graph Vector Space without Symbolic Space Guidance Summary Recent study knowledge base completion task recovering missing fact based observed fact demonstrate importance learning embeddings multistep relation Due size knowledge base previous work manually design relation path observed triplet...
[0.04855581000447273, 0.017884038388729095, 0.007883403450250626, 0.032781388610601425, -0.015366907231509686, 0.02024826966226101, -0.04147467762231827, 0.01685851253569126, 0.051046304404735565, -0.012360705062747002, 0.06187160313129425, 0.05598045140504837, -0.023612357676029205, -0.0019889480900019407, 0.012480015...
1,628
1,628
['Hongjie Shi', 'Takashi Ushio', 'Mitsuru Endo', 'Katsuyoshi Yamagami', 'Noriaki Horii']
1701.06247v1
The fifth Dialog State Tracking Challenge (DSTC5) introduces a new cross-language dialog state tracking scenario, where the participants are asked to build their trackers based on the English training corpus, while evaluating them with the unlabeled Chinese corpus. Although the computer-generated translations for both ...
A Multichannel Convolutional Neural Network For Cross-language Dialog State Tracking
2,017
http://arxiv.org/pdf/1701.06247v1
Title Multichannel Convolutional Neural Network Crosslanguage Dialog State Tracking Summary fifth Dialog State Tracking Challenge DSTC5 introduces new crosslanguage dialog state tracking scenario participant asked build tracker based English training corpus evaluating unlabeled Chinese corpus Although computergenerated...
[0.03335984796285629, 0.08617806434631348, -0.005317030940204859, 0.06914869695901871, -0.0012718377402052283, -0.009447760879993439, 0.06270889192819595, 0.02002430334687233, 0.005732350051403046, -0.06451600790023804, -0.03727588430047035, -0.07190191745758057, -0.020858945325016975, 0.06831705570220947, 0.0243591815...
1,629
1,629
['Marco Baroni', 'Armand Joulin', 'Allan Jabri', 'Germàn Kruszewski', 'Angeliki Lazaridou', 'Klemen Simonic', 'Tomas Mikolov']
1701.08954v2
With machine learning successfully applied to new daunting problems almost every day, general AI starts looking like an attainable goal. However, most current research focuses instead on important but narrow applications, such as image classification or machine translation. We believe this to be largely due to the lack...
CommAI: Evaluating the first steps towards a useful general AI
2,017
http://arxiv.org/pdf/1701.08954v2
Title CommAI Evaluating first step towards useful general AI Summary machine learning successfully applied new daunting problem almost every day general AI start looking like attainable goal However current research focus instead important narrow application image classification machine translation believe largely due ...
[0.05938199162483215, 0.03390379250049591, -0.025235941633582115, 0.0054704309441149235, -0.012722431682050228, 0.03262994438409805, 0.03638043999671936, 0.04524447023868561, -0.004807299468666315, -0.040066465735435486, 0.014416268095374107, -0.027324452996253967, 0.019468864426016808, 0.06666628271341324, 0.023373432...
1,630
1,630
['Grzegorz Chrupała', 'Lieke Gelderloos', 'Afra Alishahi']
1702.01991v3
We present a visually grounded model of speech perception which projects spoken utterances and images to a joint semantic space. We use a multi-layer recurrent highway network to model the temporal nature of spoken speech, and show that it learns to extract both form and meaning-based linguistic knowledge from the inpu...
Representations of language in a model of visually grounded speech signal
2,017
http://arxiv.org/pdf/1702.01991v3
Title Representations language model visually grounded speech signal Summary present visually grounded model speech perception project spoken utterance image joint semantic space use multilayer recurrent highway network model temporal nature spoken speech show learns extract form meaningbased linguistic knowledge input...
[-0.017416270449757576, 0.020195137709379196, -4.558585351333022e-05, 0.0636376142501831, -0.018752245232462883, -0.020176636055111885, -0.009729910641908646, 0.006369559559971094, -0.07364188879728317, -0.06754852086305618, -0.04454420134425163, 0.0015549642266705632, 0.03350938856601715, 0.07232694327831268, 0.009428...
1,631
1,631
['Tong Che', 'Yanran Li', 'Ruixiang Zhang', 'R Devon Hjelm', 'Wenjie Li', 'Yangqiu Song', 'Yoshua Bengio']
1702.07983v1
Despite the successes in capturing continuous distributions, the application of generative adversarial networks (GANs) to discrete settings, like natural language tasks, is rather restricted. The fundamental reason is the difficulty of back-propagation through discrete random variables combined with the inherent instab...
Maximum-Likelihood Augmented Discrete Generative Adversarial Networks
2,017
http://arxiv.org/pdf/1702.07983v1
Title MaximumLikelihood Augmented Discrete Generative Adversarial Networks Summary Despite success capturing continuous distribution application generative adversarial network GANs discrete setting like natural language task rather restricted fundamental reason difficulty backpropagation discrete random variable combin...
[0.03134404122829437, 0.08767537772655487, 0.005024896934628487, 0.03577417507767677, 0.013115917332470417, -0.008843081071972847, -0.010067741386592388, 0.005670683458447456, 0.014088083989918232, -0.02786160632967949, -0.005895806010812521, -0.008704510517418385, -0.005240053404122591, 0.010028756223618984, 0.0872692...
1,632
1,632
['Sercan O. Arik', 'Markus Kliegl', 'Rewon Child', 'Joel Hestness', 'Andrew Gibiansky', 'Chris Fougner', 'Ryan Prenger', 'Adam Coates']
1703.05390v3
Keyword spotting (KWS) constitutes a major component of human-technology interfaces. Maximizing the detection accuracy at a low false alarm (FA) rate, while minimizing the footprint size, latency and complexity are the goals for KWS. Towards achieving them, we study Convolutional Recurrent Neural Networks (CRNNs). Insp...
Convolutional Recurrent Neural Networks for Small-Footprint Keyword Spotting
2,017
http://arxiv.org/pdf/1703.05390v3
Title Convolutional Recurrent Neural Networks SmallFootprint Keyword Spotting Summary Keyword spotting KWS constitutes major component humantechnology interface Maximizing detection accuracy low false alarm FA rate minimizing footprint size latency complexity goal KWS Towards achieving study Convolutional Recurrent Neu...
[0.007585969753563404, 0.006239660549908876, 0.01565006747841835, 0.09081029891967773, 0.011123878881335258, -0.01505136489868164, 0.006356948520988226, 0.022448182106018066, -0.021080804988741875, -0.016063200309872627, -0.02402399852871895, -0.02390201948583126, 0.04333207756280899, 0.07359366863965988, 0.00934396497...
1,633
1,633
['Xiujun Li', 'Yun-Nung Chen', 'Lihong Li', 'Jianfeng Gao', 'Asli Celikyilmaz']
1703.07055v1
Language understanding is a key component in a spoken dialogue system. In this paper, we investigate how the language understanding module influences the dialogue system performance by conducting a series of systematic experiments on a task-oriented neural dialogue system in a reinforcement learning based setting. The ...
Investigation of Language Understanding Impact for Reinforcement Learning Based Dialogue Systems
2,017
http://arxiv.org/pdf/1703.07055v1
Title Investigation Language Understanding Impact Reinforcement Learning Based Dialogue Systems Summary Language understanding key component spoken dialogue system paper investigate language understanding module influence dialogue system performance conducting series systematic experiment taskoriented neural dialogue s...
[0.07332177460193634, -0.008202286437153816, -0.01983998343348503, 0.041516222059726715, 0.008967765606939793, 0.02742929570376873, 0.006325912196189165, -0.004597691353410482, -0.026953576132655144, -0.04791596159338951, -0.02694360353052616, 0.004234860651195049, 0.025234049186110497, 0.08306336402893066, 0.012253623...
1,634
1,634
['Baolin Peng', 'Xiujun Li', 'Lihong Li', 'Jianfeng Gao', 'Asli Celikyilmaz', 'Sungjin Lee', 'Kam-Fai Wong']
1704.03084v3
Building a dialogue agent to fulfill complex tasks, such as travel planning, is challenging because the agent has to learn to collectively complete multiple subtasks. For example, the agent needs to reserve a hotel and book a flight so that there leaves enough time for commute between arrival and hotel check-in. This p...
Composite Task-Completion Dialogue Policy Learning via Hierarchical Deep Reinforcement Learning
2,017
http://arxiv.org/pdf/1704.03084v3
Title Composite TaskCompletion Dialogue Policy Learning via Hierarchical Deep Reinforcement Learning Summary Building dialogue agent fulfill complex task travel planning challenging agent learn collectively complete multiple subtasks example agent need reserve hotel book flight leaf enough time commute arrival hotel ch...
[0.03948250785470009, 0.06048300117254257, -0.01800507679581642, -0.010744569823145866, -0.009669926948845387, 0.002322445623576641, 0.020122282207012177, -0.016591867431998253, -0.03519969433546066, -0.03942485526204109, -0.01758096180856228, -0.011416071094572544, -0.014846059493720531, 0.10274630039930344, -0.001037...
1,635
1,635
['Phong Le', 'Ivan Titov']
1704.04451v3
Coreference evaluation metrics are hard to optimize directly as they are non-differentiable functions, not easily decomposable into elementary decisions. Consequently, most approaches optimize objectives only indirectly related to the end goal, resulting in suboptimal performance. Instead, we propose a differentiable r...
Optimizing Differentiable Relaxations of Coreference Evaluation Metrics
2,017
http://arxiv.org/pdf/1704.04451v3
Title Optimizing Differentiable Relaxations Coreference Evaluation Metrics Summary Coreference evaluation metric hard optimize directly nondifferentiable function easily decomposable elementary decision Consequently approach optimize objective indirectly related end goal resulting suboptimal performance Instead propose...
[0.031887564808130264, 0.022713029757142067, 0.001530940062366426, 0.018959619104862213, -0.024802129715681076, -0.026585504412651062, -0.03582601994276047, 0.02468610554933548, -0.027281640097498894, -0.004602602683007717, -0.0227389894425869, 0.017161330208182335, 0.01625773124396801, -0.04355853050947189, -0.0113947...
1,636
1,636
['Mikhail Khodak', 'Nikunj Saunshi', 'Kiran Vodrahalli']
1704.05579v3
We introduce the Self-Annotated Reddit Corpus (SARC), a large corpus for sarcasm research and for training and evaluating systems for sarcasm detection. The corpus has 1.3 million sarcastic statements -- 10 times more than any previous dataset -- and many times more instances of non-sarcastic statements, allowing for l...
A Large Self-Annotated Corpus for Sarcasm
2,017
http://arxiv.org/pdf/1704.05579v3
Title Large SelfAnnotated Corpus Sarcasm Summary introduce SelfAnnotated Reddit Corpus SARC large corpus sarcasm research training evaluating system sarcasm detection corpus 13 million sarcastic statement 10 time previous dataset many time instance nonsarcastic statement allowing learning regime balanced unbalanced lab...
[0.09555725753307343, 0.01440010592341423, -0.0426323339343071, 0.031194519251585007, -0.02130887471139431, 0.02186202071607113, 0.012995390221476555, 0.029476240277290344, 0.007560833357274532, -0.04443804919719696, -0.020089365541934967, 0.06845533847808838, -0.03160558268427849, -0.0016147756250575185, -0.0268300343...
1,637
1,637
['Qizhe Xie', 'Xuezhe Ma', 'Zihang Dai', 'Eduard Hovy']
1704.05908v2
Knowledge bases are important resources for a variety of natural language processing tasks but suffer from incompleteness. We propose a novel embedding model, \emph{ITransF}, to perform knowledge base completion. Equipped with a sparse attention mechanism, ITransF discovers hidden concepts of relations and transfer sta...
An Interpretable Knowledge Transfer Model for Knowledge Base Completion
2,017
http://arxiv.org/pdf/1704.05908v2
Title Interpretable Knowledge Transfer Model Knowledge Base Completion Summary Knowledge base important resource variety natural language processing task suffer incompleteness propose novel embedding model emphITransF perform knowledge base completion Equipped sparse attention mechanism ITransF discovers hidden concept...
[0.043073348701000214, 0.005817985162138939, -0.012169345282018185, 0.0677141398191452, -0.030289538204669952, -0.009101198986172676, -0.07530991733074188, 0.01479955855756998, -0.007260593120008707, -0.05178390070796013, -0.020985275506973267, 0.04752083495259285, 0.021122770383954048, 0.04829082638025284, 0.033371694...
1,638
1,638
['Sida I. Wang', 'Samuel Ginn', 'Percy Liang', 'Christoper D. Manning']
1704.06956v1
Our goal is to create a convenient natural language interface for performing well-specified but complex actions such as analyzing data, manipulating text, and querying databases. However, existing natural language interfaces for such tasks are quite primitive compared to the power one wields with a programming language...
Naturalizing a Programming Language via Interactive Learning
2,017
http://arxiv.org/pdf/1704.06956v1
Title Naturalizing Programming Language via Interactive Learning Summary goal create convenient natural language interface performing wellspecified complex action analyzing data manipulating text querying database However existing natural language interface task quite primitive compared power one wields programming lan...
[0.010968358255922794, 0.04067239537835121, -0.03744404762983322, 0.017588863149285316, -0.007274262607097626, 0.02617216669023037, -0.04348306730389595, 0.001065506599843502, 0.009838555008172989, -0.007130177225917578, 0.0008666044450365007, 0.04897729679942131, 0.008853831328451633, 0.08185174316167831, 0.0105252442...
1,639
1,639
['Ganbin Zhou', 'Ping Luo', 'Rongyu Cao', 'Yijun Xiao', 'Fen Lin', 'Bo Chen', 'Qing He']
1705.00321v4
Different from other sequential data, sentences in natural language are structured by linguistic grammars. Previous generative conversational models with chain-structured decoder ignore this structure in human language and might generate plausible responses with less satisfactory relevance and fluency. In this study, w...
Tree-Structured Neural Machine for Linguistics-Aware Sentence Generation
2,017
http://arxiv.org/pdf/1705.00321v4
Title TreeStructured Neural Machine LinguisticsAware Sentence Generation Summary Different sequential data sentence natural language structured linguistic grammar Previous generative conversational model chainstructured decoder ignore structure human language might generate plausible response le satisfactory relevance ...
[0.07612986117601395, 0.05658919736742973, -0.008598160929977894, 0.044696543365716934, -0.05268942937254906, -0.01184448879212141, -0.04538199305534363, -0.0009932274697348475, -0.022164301946759224, -0.07526205480098724, 0.024471236392855644, 0.0017190936487168074, 0.026093056425452232, 0.04956630617380142, -0.015354...
1,640
1,640
['Hanxiao Liu', 'Yuexin Wu', 'Yiming Yang']
1705.02426v2
Large-scale multi-relational embedding refers to the task of learning the latent representations for entities and relations in large knowledge graphs. An effective and scalable solution for this problem is crucial for the true success of knowledge-based inference in a broad range of applications. This paper proposes a ...
Analogical Inference for Multi-Relational Embeddings
2,017
http://arxiv.org/pdf/1705.02426v2
Title Analogical Inference MultiRelational Embeddings Summary Largescale multirelational embedding refers task learning latent representation entity relation large knowledge graph effective scalable solution problem crucial true success knowledgebased inference broad range application paper proposes novel framework opt...
[0.01589697040617466, 0.03808876499533653, -0.006677874829620123, 0.06699118763208389, 0.012618601322174072, 0.0325901061296463, 0.01929960399866104, -0.0010768149513751268, 0.029928559437394142, -0.04898905009031296, -0.0018525781342759728, 0.044291067868471146, -0.045193735510110855, -0.011963223107159138, 0.04813208...
1,641
1,641
['Wang Ling', 'Dani Yogatama', 'Chris Dyer', 'Phil Blunsom']
1705.04146v3
Solving algebraic word problems requires executing a series of arithmetic operations---a program---to obtain a final answer. However, since programs can be arbitrarily complicated, inducing them directly from question-answer pairs is a formidable challenge. To make this task more feasible, we solve these problems by ge...
Program Induction by Rationale Generation : Learning to Solve and Explain Algebraic Word Problems
2,017
http://arxiv.org/pdf/1705.04146v3
Title Program Induction Rationale Generation Learning Solve Explain Algebraic Word Problems Summary Solving algebraic word problem requires executing series arithmetic operationsa programto obtain final answer However since program arbitrarily complicated inducing directly questionanswer pair formidable challenge make ...
[-0.01704597845673561, 0.07351915538311005, -0.023795681074261665, 0.017611980438232422, -0.03423978015780449, 0.013979166746139526, 0.0020346748642623425, 0.021386021748185158, -0.04356525465846062, 0.0023559601977467537, 0.05891282856464386, 0.0404079370200634, 0.020094646140933037, 0.07591884583234787, 0.01881250180...
1,642
1,642
['Rakshit Trivedi', 'Hanjun Dai', 'Yichen Wang', 'Le Song']
1705.05742v3
The availability of large scale event data with time stamps has given rise to dynamically evolving knowledge graphs that contain temporal information for each edge. Reasoning over time in such dynamic knowledge graphs is not yet well understood. To this end, we present Know-Evolve, a novel deep evolutionary knowledge n...
Know-Evolve: Deep Temporal Reasoning for Dynamic Knowledge Graphs
2,017
http://arxiv.org/pdf/1705.05742v3
Title KnowEvolve Deep Temporal Reasoning Dynamic Knowledge Graphs Summary availability large scale event data time stamp given rise dynamically evolving knowledge graph contain temporal information edge Reasoning time dynamic knowledge graph yet well understood end present KnowEvolve novel deep evolutionary knowledge n...
[-0.014916440472006798, 0.014659629203379154, -0.021968496963381767, 0.02449040114879608, 0.004986301995813847, 0.00021810548787470907, -0.025449981912970543, 0.012270137667655945, 0.008984942920506, 0.012035991065204144, 0.08493039011955261, 0.022060047835111618, -0.04567502439022064, 0.09954413026571274, 0.0088897598...
1,643
1,643
['Jiatao Gu', 'Yong Wang', 'Kyunghyun Cho', 'Victor O. K. Li']
1705.07267v2
In this paper, we extend an attention-based neural machine translation (NMT) model by allowing it to access an entire training set of parallel sentence pairs even after training. The proposed approach consists of two stages. In the first stage--retrieval stage--, an off-the-shelf, black-box search engine is used to ret...
Search Engine Guided Non-Parametric Neural Machine Translation
2,017
http://arxiv.org/pdf/1705.07267v2
Title Search Engine Guided NonParametric Neural Machine Translation Summary paper extend attentionbased neural machine translation NMT model allowing access entire training set parallel sentence pair even training proposed approach consists two stage first stageretrieval stage offtheshelf blackbox search engine used re...
[0.04754798486828804, 0.028966352343559265, -0.01885300688445568, 0.0394458994269371, -0.04259197786450386, -0.005891206208616495, -0.014658703468739986, -0.005972510203719139, -0.0279019046574831, -0.03900972753763199, -0.05010853707790375, -0.01946268603205681, -0.0113054309040308, -0.01740819588303566, 0.04763027280...
1,644
1,644
['James Foulds']
1705.07368v3
Word embeddings improve the performance of NLP systems by revealing the hidden structural relationships between words. Despite their success in many applications, word embeddings have seen very little use in computational social science NLP tasks, presumably due to their reliance on big data, and to a lack of interpret...
Mixed Membership Word Embeddings for Computational Social Science
2,017
http://arxiv.org/pdf/1705.07368v3
Title Mixed Membership Word Embeddings Computational Social Science Summary Word embeddings improve performance NLP system revealing hidden structural relationship word Despite success many application word embeddings seen little use computational social science NLP task presumably due reliance big data lack interpreta...
[0.01931675709784031, 0.04177436605095863, -0.02943039871752262, 0.06103846803307533, 0.007655159570276737, 0.027545196935534477, -0.02873079478740692, -0.0005909066530875862, 0.02217060513794422, -0.08300824463367462, 0.008766704238951206, -0.012200575321912766, 0.02606239728629589, 0.05520587041974068, 0.001563336583...
1,645
1,645
['Yishu Miao', 'Edward Grefenstette', 'Phil Blunsom']
1706.00359v1
Topic models have been widely explored as probabilistic generative models of documents. Traditional inference methods have sought closed-form derivations for updating the models, however as the expressiveness of these models grows, so does the difficulty of performing fast and accurate inference over their parameters. ...
Discovering Discrete Latent Topics with Neural Variational Inference
2,017
http://arxiv.org/pdf/1706.00359v1
Title Discovering Discrete Latent Topics Neural Variational Inference Summary Topic model widely explored probabilistic generative model document Traditional inference method sought closedform derivation updating model however expressiveness model grows difficulty performing fast accurate inference parameter paper pres...
[0.04606425389647484, 0.06148383021354675, -0.028130831196904182, 0.01759776473045349, -0.014205140061676502, -0.028366142883896828, -0.010045989416539669, -0.02873123064637184, -0.06671158969402313, 0.0077750543132424355, 0.024553263559937477, -0.005977964494377375, 0.000892955984454602, 0.038950905203819275, 0.029719...
1,646
1,646
['Omer Levy', 'Minjoon Seo', 'Eunsol Choi', 'Luke Zettlemoyer']
1706.04115v1
We show that relation extraction can be reduced to answering simple reading comprehension questions, by associating one or more natural-language questions with each relation slot. This reduction has several advantages: we can (1) learn relation-extraction models by extending recent neural reading-comprehension techniqu...
Zero-Shot Relation Extraction via Reading Comprehension
2,017
http://arxiv.org/pdf/1706.04115v1
Title ZeroShot Relation Extraction via Reading Comprehension Summary show relation extraction reduced answering simple reading comprehension question associating one naturallanguage question relation slot reduction several advantage 1 learn relationextraction model extending recent neural readingcomprehension technique...
[0.046090006828308105, 0.029463212937116623, 0.0077476114965975285, 0.10946109890937805, -0.020484905689954758, 0.013023667968809605, 0.008800984360277653, 0.025328950956463814, 0.018231919035315514, -0.019706858322024345, -0.02056054025888443, 0.005878789816051722, 0.006001424044370651, -0.01030429545789957, -0.001378...
1,647
1,647
['Suncong Zheng', 'Feng Wang', 'Hongyun Bao', 'Yuexing Hao', 'Peng Zhou', 'Bo Xu']
1706.05075v1
Joint extraction of entities and relations is an important task in information extraction. To tackle this problem, we firstly propose a novel tagging scheme that can convert the joint extraction task to a tagging problem. Then, based on our tagging scheme, we study different end-to-end models to extract entities and th...
Joint Extraction of Entities and Relations Based on a Novel Tagging Scheme
2,017
http://arxiv.org/pdf/1706.05075v1
Title Joint Extraction Entities Relations Based Novel Tagging Scheme Summary Joint extraction entity relation important task information extraction tackle problem firstly propose novel tagging scheme convert joint extraction task tagging problem based tagging scheme study different endtoend model extract entity relatio...
[0.0449993833899498, 0.0188040342181921, -0.005307162646204233, 0.036217521876096725, -0.016013767570257187, 0.04766160994768143, -0.007558783516287804, 0.06348273158073425, 0.056852489709854126, -0.051802244037389755, -0.01499092485755682, 0.0331961065530777, -0.018101878464221954, 0.007167470175772905, -0.02759459428...
1,648
1,648
['Devendra Singh Chaplot', 'Kanthashree Mysore Sathyendra', 'Rama Kumar Pasumarthi', 'Dheeraj Rajagopal', 'Ruslan Salakhutdinov']
1706.07230v2
To perform tasks specified by natural language instructions, autonomous agents need to extract semantically meaningful representations of language and map it to visual elements and actions in the environment. This problem is called task-oriented language grounding. We propose an end-to-end trainable neural architecture...
Gated-Attention Architectures for Task-Oriented Language Grounding
2,017
http://arxiv.org/pdf/1706.07230v2
Title GatedAttention Architectures TaskOriented Language Grounding Summary perform task specified natural language instruction autonomous agent need extract semantically meaningful representation language map visual element action environment problem called taskoriented language grounding propose endtoend trainable neu...
[0.049830831587314606, -0.028533989563584328, -0.012531942687928677, 0.057402320206165314, -0.013469807803630829, -0.0028702090494334698, 0.01973002776503563, -0.02122444286942482, 0.006750267930328846, -0.0561261922121048, -0.05564664676785469, 0.01617429219186306, -0.012247184291481972, 0.07842019200325012, 0.0418636...
1,649
1,649
['Michael Janner', 'Karthik Narasimhan', 'Regina Barzilay']
1707.03938v2
The interpretation of spatial references is highly contextual, requiring joint inference over both language and the environment. We consider the task of spatial reasoning in a simulated environment, where an agent can act and receive rewards. The proposed model learns a representation of the world steered by instructio...
Representation Learning for Grounded Spatial Reasoning
2,017
http://arxiv.org/pdf/1707.03938v2
Title Representation Learning Grounded Spatial Reasoning Summary interpretation spatial reference highly contextual requiring joint inference language environment consider task spatial reasoning simulated environment agent act receive reward proposed model learns representation world steered instruction text design all...
[0.05973513796925545, -0.004765619058161974, -0.011242805048823357, 0.0004944343236275017, -0.03505253046751022, 0.0028340753633528948, 0.033819589763879776, -0.037656769156455994, -0.0032195316161960363, -0.05353553220629692, -0.017773672938346863, 0.050715360790491104, 0.017630312591791153, 0.033936142921447754, 0.01...
1,650
1,650
['Debanjan Ghosh', 'Alexander Richard Fabbri', 'Smaranda Muresan']
1707.06226v1
Computational models for sarcasm detection have often relied on the content of utterances in isolation. However, speaker's sarcastic intent is not always obvious without additional context. Focusing on social media discussions, we investigate two issues: (1) does modeling of conversation context help in sarcasm detecti...
The Role of Conversation Context for Sarcasm Detection in Online Interactions
2,017
http://arxiv.org/pdf/1707.06226v1
Title Role Conversation Context Sarcasm Detection Online Interactions Summary Computational model sarcasm detection often relied content utterance isolation However speaker sarcastic intent always obvious without additional context Focusing social medium discussion investigate two issue 1 modeling conversation context ...
[0.08541964739561081, -0.022232238203287125, -0.04826135188341141, 0.050598304718732834, -0.043831199407577515, 0.01492568850517273, 0.01859159953892231, 0.020948415622115135, -0.010316837579011917, -0.052229974418878555, -0.024141333997249603, 0.002883593551814556, 0.0063470457680523396, 0.042905792593955994, -0.02324...
1,651
1,651
['Bryan McCann', 'James Bradbury', 'Caiming Xiong', 'Richard Socher']
1708.00107v1
Computer vision has benefited from initializing multiple deep layers with weights pretrained on large supervised training sets like ImageNet. Natural language processing (NLP) typically sees initialization of only the lowest layer of deep models with pretrained word vectors. In this paper, we use a deep LSTM encoder fr...
Learned in Translation: Contextualized Word Vectors
2,017
http://arxiv.org/pdf/1708.00107v1
Title Learned Translation Contextualized Word Vectors Summary Computer vision benefited initializing multiple deep layer weight pretrained large supervised training set like ImageNet Natural language processing NLP typically see initialization lowest layer deep model pretrained word vector paper use deep LSTM encoder a...
[0.049259379506111145, 0.04649924486875534, 0.007558019831776619, 0.07306995242834091, -0.013656089082360268, 0.01776052825152874, 0.03649386018514633, 0.0316699855029583, -0.03937124088406563, -0.06346237659454346, -0.010479360818862915, -0.0014433582546189427, 0.01245107315480709, 0.02059072256088257, 0.0383643619716...
1,652
1,652
['Karthik Narasimhan', 'Regina Barzilay', 'Tommi Jaakkola']
1708.00133v1
In this paper, we explore the utilization of natural language to drive transfer for reinforcement learning (RL). Despite the wide-spread application of deep RL techniques, learning generalized policy representations that work across domains remains a challenging problem. We demonstrate that textual descriptions of envi...
Deep Transfer in Reinforcement Learning by Language Grounding
2,017
http://arxiv.org/pdf/1708.00133v1
Title Deep Transfer Reinforcement Learning Language Grounding Summary paper explore utilization natural language drive transfer reinforcement learning RL Despite widespread application deep RL technique learning generalized policy representation work across domain remains challenging problem demonstrate textual descrip...
[0.043645184487104416, 0.043562207370996475, 0.0028182680252939463, 0.018314791843295097, -0.0017700043972581625, 0.003679888788610697, 0.015060151927173138, -0.03565235435962677, -0.043955136090517044, -0.04896461218595505, -0.035248614847660065, 0.011577161960303783, -0.014589316211640835, 0.05924428254365921, -0.003...
1,653
1,653
['Clemens Rosenbaum', 'Tian Gao', 'Tim Klinger']
1708.01776v1
In this paper we present a new dataset and user simulator e-QRAQ (explainable Query, Reason, and Answer Question) which tests an Agent's ability to read an ambiguous text; ask questions until it can answer a challenge question; and explain the reasoning behind its questions and answer. The User simulator provides the A...
e-QRAQ: A Multi-turn Reasoning Dataset and Simulator with Explanations
2,017
http://arxiv.org/pdf/1708.01776v1
Title eQRAQ Multiturn Reasoning Dataset Simulator Explanations Summary paper present new dataset user simulator eQRAQ explainable Query Reason Answer Question test Agents ability read ambiguous text ask question answer challenge question explain reasoning behind question answer User simulator provides Agent short ambig...
[0.02729903534054756, 0.029129182919859886, -0.03543124347925186, 0.025677554309368134, -0.0072814831510186195, 0.01694638282060623, 0.015348090790212154, -0.027709657326340675, 0.00574113056063652, -0.0015496464911848307, 0.02961283177137375, 0.01789160817861557, -0.028595812618732452, 0.054605793207883835, 0.02147383...
1,654
1,654
['Yixin Nie', 'Mohit Bansal']
1708.02312v2
We present a simple sequential sentence encoder for multi-domain natural language inference. Our encoder is based on stacked bidirectional LSTM-RNNs with shortcut connections and fine-tuning of word embeddings. The overall supervised model uses the above encoder to encode two input sentences into two vectors, and then ...
Shortcut-Stacked Sentence Encoders for Multi-Domain Inference
2,017
http://arxiv.org/pdf/1708.02312v2
Title ShortcutStacked Sentence Encoders MultiDomain Inference Summary present simple sequential sentence encoder multidomain natural language inference encoder based stacked bidirectional LSTMRNNs shortcut connection finetuning word embeddings overall supervised model us encoder encode two input sentence two vector us ...
[0.03245577588677406, 0.075023353099823, 0.00549362413585186, 0.07455085963010788, -0.0376153327524662, 0.03223928436636925, 0.028135372325778008, -0.03679712489247322, -0.019055459648370743, -0.057231102138757706, -0.039579302072525024, -0.0038737214636057615, 0.005428060423582792, 0.011539352126419544, 0.054722089320...
1,655
1,655
['Meng Fang', 'Yuan Li', 'Trevor Cohn']
1708.02383v1
Active learning aims to select a small subset of data for annotation such that a classifier learned on the data is highly accurate. This is usually done using heuristic selection methods, however the effectiveness of such methods is limited and moreover, the performance of heuristics varies between datasets. To address...
Learning how to Active Learn: A Deep Reinforcement Learning Approach
2,017
http://arxiv.org/pdf/1708.02383v1
Title Learning Active Learn Deep Reinforcement Learning Approach Summary Active learning aim select small subset data annotation classifier learned data highly accurate usually done using heuristic selection method however effectiveness method limited moreover performance heuristic varies datasets address shortcoming i...
[0.01804141141474247, 0.007563535589724779, -0.006647841539233923, -0.004702047910541296, 0.012363075278699398, 0.054901961237192154, 0.013802614994347095, 0.014474201016128063, 0.02405369095504284, -0.0332019105553627, -0.004608800169080496, -0.008046170696616173, -0.04925275966525078, 0.046882640570402145, -0.0007344...
1,656
1,656
['Yu Wang', 'Jiayi Liu', 'Yuxiang Liu', 'Jun Hao', 'Yang He', 'Jinghe Hu', 'Weipeng P. Yan', 'Mantian Li']
1708.05565v2
We present LADDER, the first deep reinforcement learning agent that can successfully learn control policies for large-scale real-world problems directly from raw inputs composed of high-level semantic information. The agent is based on an asynchronous stochastic variant of DQN (Deep Q Network) named DASQN. The inputs o...
LADDER: A Human-Level Bidding Agent for Large-Scale Real-Time Online Auctions
2,017
http://arxiv.org/pdf/1708.05565v2
Title LADDER HumanLevel Bidding Agent LargeScale RealTime Online Auctions Summary present LADDER first deep reinforcement learning agent successfully learn control policy largescale realworld problem directly raw input composed highlevel semantic information agent based asynchronous stochastic variant DQN Deep Q Networ...
[0.029619039967656136, 0.0847579836845398, -0.02971203438937664, -0.023376688361167908, -0.013757727108895779, -0.013228124938905239, 0.023451605811715126, 0.01019690278917551, -0.04340572655200958, -0.025120826438069344, -0.04074287787079811, -0.003977801650762558, -0.014474058523774147, 0.13652123510837555, 0.0085918...
1,657
1,657
['Ryan Lowe', 'Michael Noseworthy', 'Iulian V. Serban', 'Nicolas Angelard-Gontier', 'Yoshua Bengio', 'Joelle Pineau']
1708.07149v2
Automatically evaluating the quality of dialogue responses for unstructured domains is a challenging problem. Unfortunately, existing automatic evaluation metrics are biased and correlate very poorly with human judgements of response quality. Yet having an accurate automatic evaluation procedure is crucial for dialogue...
Towards an Automatic Turing Test: Learning to Evaluate Dialogue Responses
2,017
http://arxiv.org/pdf/1708.07149v2
Title Towards Automatic Turing Test Learning Evaluate Dialogue Responses Summary Automatically evaluating quality dialogue response unstructured domain challenging problem Unfortunately existing automatic evaluation metric biased correlate poorly human judgement response quality Yet accurate automatic evaluation proced...
[0.059065233916044235, 0.021957209333777428, -0.011664843186736107, 0.030072690919041634, -0.004112504422664642, -0.004852978978306055, -0.0001845026999944821, 0.019366279244422913, 0.02541208267211914, -0.061102669686079025, -0.07446898519992828, -0.007445188704878092, 0.019084012135863304, 0.0788661390542984, -0.0118...
1,658
1,658
['Enrique Noriega-Atala', 'Marco A. Valenzuela-Escarcega', 'Clayton T. Morrison', 'Mihai Surdeanu']
1709.00149v1
Recent efforts in bioinformatics have achieved tremendous progress in the machine reading of biomedical literature, and the assembly of the extracted biochemical interactions into large-scale models such as protein signaling pathways. However, batch machine reading of literature at today's scale (PubMed alone indexes o...
Learning what to read: Focused machine reading
2,017
http://arxiv.org/pdf/1709.00149v1
Title Learning read Focused machine reading Summary Recent effort bioinformatics achieved tremendous progress machine reading biomedical literature assembly extracted biochemical interaction largescale model protein signaling pathway However batch machine reading literature today scale PubMed alone index 1 million pape...
[0.061723433434963226, 0.031000150367617607, 0.01346401497721672, -0.02260061912238598, -0.016054661944508553, 0.02737303078174591, 0.027434850111603737, 0.04379849508404732, 0.01398156676441431, -0.010996649973094463, 0.013167965225875378, 0.013666392304003239, -0.0030092617962509394, 0.07026710361242294, -0.024531522...
1,659
1,659
['Wen Zhang', 'Jiawei Hu', 'Yang Feng', 'Qun Liu']
1709.03980v2
Although neural machine translation (NMT) with the encoder-decoder framework has achieved great success in recent times, it still suffers from some drawbacks: RNNs tend to forget old information which is often useful and the encoder only operates through words without considering word relationship. To solve these probl...
Refining Source Representations with Relation Networks for Neural Machine Translation
2,017
http://arxiv.org/pdf/1709.03980v2
Title Refining Source Representations Relation Networks Neural Machine Translation Summary Although neural machine translation NMT encoderdecoder framework achieved great success recent time still suffers drawback RNNs tend forget old information often useful encoder operates word without considering word relationship ...
[0.0354117751121521, 0.0869050845503807, 0.004381262231618166, 0.07732675224542618, -0.024843480437994003, -0.008139606565237045, -0.012364808470010757, 0.028637468814849854, -0.06733638793230057, -0.008298108354210854, -0.027959683910012245, -0.011375104077160358, 0.03084663487970829, -0.004550204612314701, -0.0037673...
1,660
1,660
['Yuyu Zhang', 'Hanjun Dai', 'Zornitsa Kozareva', 'Alexander J. Smola', 'Le Song']
1709.04071v5
Knowledge graph (KG) is known to be helpful for the task of question answering (QA), since it provides well-structured relational information between entities, and allows one to further infer indirect facts. However, it is challenging to build QA systems which can learn to reason over knowledge graphs based on question...
Variational Reasoning for Question Answering with Knowledge Graph
2,017
http://arxiv.org/pdf/1709.04071v5
Title Variational Reasoning Question Answering Knowledge Graph Summary Knowledge graph KG known helpful task question answering QA since provides wellstructured relational information entity allows one infer indirect fact However challenging build QA system learn reason knowledge graph based questionanswer pair alone F...
[0.04944609850645065, 0.08523157238960266, -0.004002944100648165, 0.021603383123874664, -0.030366288498044014, 0.010183987207710743, 0.009923948906362057, 0.007790801115334034, -0.006559575907886028, 0.0013447346864268184, 0.014538264833390713, -0.0187444519251585, -0.0011585148749873042, 0.061367906630039215, 0.042693...
1,661
1,661
['Kamran Kowsari', 'Donald E. Brown', 'Mojtaba Heidarysafa', 'Kiana Jafari Meimandi', 'Matthew S. Gerber', 'Laura E. Barnes']
1709.08267v2
The continually increasing number of documents produced each year necessitates ever improving information processing methods for searching, retrieving, and organizing text. Central to these information processing methods is document classification, which has become an important application for supervised learning. Rece...
HDLTex: Hierarchical Deep Learning for Text Classification
2,017
http://arxiv.org/pdf/1709.08267v2
Title HDLTex Hierarchical Deep Learning Text Classification Summary continually increasing number document produced year necessitates ever improving information processing method searching retrieving organizing text Central information processing method document classification become important application supervised le...
[0.029860256239771843, 0.05068176984786987, -0.0031309635378420353, 0.011257732287049294, -0.026909852400422096, 0.03616732731461525, 0.06710837036371231, 0.036226946860551834, 0.007124762516468763, -0.073305144906044, -0.026329921558499336, -0.001008965540677309, -0.01977689564228058, 0.04175018146634102, -0.022344693...
1,662
1,662
['Jiaxian Guo', 'Sidi Lu', 'Han Cai', 'Weinan Zhang', 'Yong Yu', 'Jun Wang']
1709.08624v2
Automatically generating coherent and semantically meaningful text has many applications in machine translation, dialogue systems, image captioning, etc. Recently, by combining with policy gradient, Generative Adversarial Nets (GAN) that use a discriminative model to guide the training of the generative model as a rein...
Long Text Generation via Adversarial Training with Leaked Information
2,017
http://arxiv.org/pdf/1709.08624v2
Title Long Text Generation via Adversarial Training Leaked Information Summary Automatically generating coherent semantically meaningful text many application machine translation dialogue system image captioning etc Recently combining policy gradient Generative Adversarial Nets GAN use discriminative model guide traini...
[0.044151391834020615, 0.06765884160995483, 0.000773488893173635, 0.0353635735809803, -0.008926407434046268, -0.024775858968496323, 0.029922906309366226, 0.024829015135765076, -0.006913799326866865, -0.016501082107424736, -0.015357259660959244, -0.008059680461883545, 0.00457238033413887, 0.06717986613512039, 0.03766176...
1,663
1,663
['Yanchao Yu', 'Arash Eshghi', 'Oliver Lemon']
1709.10426v1
We present a multi-modal dialogue system for interactive learning of perceptually grounded word meanings from a human tutor. The system integrates an incremental, semantic parsing/generation framework - Dynamic Syntax and Type Theory with Records (DS-TTR) - with a set of visual classifiers that are learned throughout t...
Training an adaptive dialogue policy for interactive learning of visually grounded word meanings
2,017
http://arxiv.org/pdf/1709.10426v1
Title Training adaptive dialogue policy interactive learning visually grounded word meaning Summary present multimodal dialogue system interactive learning perceptually grounded word meaning human tutor system integrates incremental semantic parsinggeneration framework Dynamic Syntax Type Theory Records DSTTR set visua...
[0.07690167427062988, -0.0007830485119484365, -0.019641989842057228, 0.04991159960627556, -0.006082721054553986, -0.005488160997629166, 0.02760549634695053, -0.0022407355718314648, -0.029501063749194145, -0.07111825793981552, -0.020298799499869347, 0.030499771237373352, 0.00606605876237154, 0.06865528970956802, 0.01282...
1,664
1,664
['Yanchao Yu', 'Arash Eshghi', 'Gregory Mills', 'Oliver Joseph Lemon']
1709.10431v1
We motivate and describe a new freely available human-human dialogue dataset for interactive learning of visually grounded word meanings through ostensive definition by a tutor to a learner. The data has been collected using a novel, character-by-character variant of the DiET chat tool (Healey et al., 2003; Mills and H...
The BURCHAK corpus: a Challenge Data Set for Interactive Learning of Visually Grounded Word Meanings
2,017
http://arxiv.org/pdf/1709.10431v1
Title BURCHAK corpus Challenge Data Set Interactive Learning Visually Grounded Word Meanings Summary motivate describe new freely available humanhuman dialogue dataset interactive learning visually grounded word meaning ostensive definition tutor learner data collected using novel characterbycharacter variant DiET chat...
[0.07391994446516037, 0.029208306223154068, -0.019237523898482323, 0.018998458981513977, -0.02345430850982666, -0.0051526762545108795, -0.006739665754139423, 0.038098596036434174, 0.00595109025016427, -0.04847061261534691, -0.03365464136004448, 0.004675155505537987, 0.03135478124022484, 0.07411491125822067, 0.008221687...
1,665
1,665
['Mikel Artetxe', 'Gorka Labaka', 'Eneko Agirre', 'Kyunghyun Cho']
1710.11041v2
In spite of the recent success of neural machine translation (NMT) in standard benchmarks, the lack of large parallel corpora poses a major practical problem for many language pairs. There have been several proposals to alleviate this issue with, for instance, triangulation and semi-supervised learning techniques, but ...
Unsupervised Neural Machine Translation
2,017
http://arxiv.org/pdf/1710.11041v2
Title Unsupervised Neural Machine Translation Summary spite recent success neural machine translation NMT standard benchmark lack large parallel corpus pose major practical problem many language pair several proposal alleviate issue instance triangulation semisupervised learning technique still require strong crossling...
[0.01575191132724285, 0.029353687539696693, 0.006226384546607733, 0.0696263238787651, -0.046245645731687546, 0.01894947700202465, 0.02305709943175316, -0.014726054854691029, -0.002702227095142007, -0.019429173320531845, -0.024393292143940926, 0.0050637549720704556, 0.02559409663081169, -0.019150953739881516, 0.04965849...
1,666
1,666
['Baolin Peng', 'Xiujun Li', 'Jianfeng Gao', 'Jingjing Liu', 'Yun-Nung Chen', 'Kam-Fai Wong']
1710.11277v2
This paper presents a new method --- adversarial advantage actor-critic (Adversarial A2C), which significantly improves the efficiency of dialogue policy learning in task-completion dialogue systems. Inspired by generative adversarial networks (GAN), we train a discriminator to differentiate responses/actions generated...
Adversarial Advantage Actor-Critic Model for Task-Completion Dialogue Policy Learning
2,017
http://arxiv.org/pdf/1710.11277v2
Title Adversarial Advantage ActorCritic Model TaskCompletion Dialogue Policy Learning Summary paper present new method adversarial advantage actorcritic Adversarial A2C significantly improves efficiency dialogue policy learning taskcompletion dialogue system Inspired generative adversarial network GAN train discriminat...
[0.04921393468976021, 0.0812935084104538, -0.02069556526839733, 0.01552529912441969, 0.009415358304977417, -0.0008003063267096877, 0.013610521331429482, -0.02162131853401661, -0.0051413350738584995, -0.013016744516789913, -0.010101469233632088, -0.010248959995806217, -0.0221969336271286, 0.06796622276306152, 0.02855882...
1,667
1,667
['Anselm Rothe', 'Brenden M. Lake', 'Todd M. Gureckis']
1711.06351v1
A hallmark of human intelligence is the ability to ask rich, creative, and revealing questions. Here we introduce a cognitive model capable of constructing human-like questions. Our approach treats questions as formal programs that, when executed on the state of the world, output an answer. The model specifies a probab...
Question Asking as Program Generation
2,017
http://arxiv.org/pdf/1711.06351v1
Title Question Asking Program Generation Summary hallmark human intelligence ability ask rich creative revealing question introduce cognitive model capable constructing humanlike question approach treat question formal program executed state world output answer model specifies probability distribution complex compositi...
[0.0700138658285141, 0.010607939213514328, -0.03699873387813568, -0.006614347454160452, -0.022612515836954117, 0.007586480118334293, 0.009983036667108536, -0.0001534219045424834, 0.00286739319562912, 0.000804007111582905, 0.03503533452749252, 0.006783115211874247, -0.01565917395055294, 0.09400889277458191, 0.0288667492...
1,668
1,668
['Xiaopeng Yang', 'Xiaowen Lin', 'Shunda Suo', 'Ming Li']
1711.07632v2
Computer poetry generation is our first step towards computer writing. Writing must have a theme. The current approaches of using sequence-to-sequence models with attention often produce non-thematic poems. We present a novel conditional variational autoencoder with a hybrid decoder adding the deconvolutional neural ne...
Generating Thematic Chinese Poetry using Conditional Variational Autoencoders with Hybrid Decoders
2,017
http://arxiv.org/pdf/1711.07632v2
Title Generating Thematic Chinese Poetry using Conditional Variational Autoencoders Hybrid Decoders Summary Computer poetry generation first step towards computer writing Writing must theme current approach using sequencetosequence model attention often produce nonthematic poem present novel conditional variational aut...
[0.07178948074579239, 0.08589985221624374, -0.02554338052868843, 0.03470619395375252, -0.013450779020786285, -0.00901603139936924, 0.009006761945784092, -0.028855247423052788, -0.03300664946436882, 6.183225195854902e-05, 0.045623961836099625, -0.03440355136990547, 0.02182793617248535, 0.051854681223630905, -0.011835519...
1,669
1,669
['Igor Melnyk', 'Cicero Nogueira dos Santos', 'Kahini Wadhawan', 'Inkit Padhi', 'Abhishek Kumar']
1711.09395v2
Text attribute transfer using non-parallel data requires methods that can perform disentanglement of content and linguistic attributes. In this work, we propose multiple improvements over the existing approaches that enable the encoder-decoder framework to cope with the text attribute transfer from non-parallel data. W...
Improved Neural Text Attribute Transfer with Non-parallel Data
2,017
http://arxiv.org/pdf/1711.09395v2
Title Improved Neural Text Attribute Transfer Nonparallel Data Summary Text attribute transfer using nonparallel data requires method perform disentanglement content linguistic attribute work propose multiple improvement existing approach enable encoderdecoder framework cope text attribute transfer nonparallel data per...
[-0.017598232254385948, 0.12090269476175308, 0.005992837715893984, 0.038360536098480225, -0.030940232798457146, 0.008306553587317467, -0.03400611877441406, 0.011053790338337421, -0.0450875386595726, -0.020811334252357483, -0.05254143849015236, 0.015640124678611755, 0.03403935208916664, 0.04989570379257202, -0.005646382...
1,670
1,670
['Xu Sun', 'Weiwei Sun', 'Shuming Ma', 'Xuancheng Ren', 'Yi Zhang', 'Wenjie Li', 'Houfeng Wang']
1711.10331v1
Recent systems on structured prediction focus on increasing the level of structural dependencies within the model. However, our study suggests that complex structures entail high overfitting risks. To control the structure-based overfitting, we propose to conduct structure regularization decoding (SR decoding). The dec...
Complex Structure Leads to Overfitting: A Structure Regularization Decoding Method for Natural Language Processing
2,017
http://arxiv.org/pdf/1711.10331v1
Title Complex Structure Leads Overfitting Structure Regularization Decoding Method Natural Language Processing Summary Recent system structured prediction focus increasing level structural dependency within model However study suggests complex structure entail high overfitting risk control structurebased overfitting pr...
[0.03847465664148331, 0.01903417333960533, 0.007986175827682018, 0.03570616990327835, 0.00906192883849144, 0.020042255520820618, -0.0457451269030571, 0.007653772830963135, -0.0009377449750900269, -0.02469203807413578, -0.017398443073034286, -0.021035166457295418, 0.024300744757056236, 0.04207364469766617, -0.0147826746...
1,671
1,671
['Arthur Bražinskas', 'Serhii Havrylov', 'Ivan Titov']
1711.11027v1
We introduce a method for embedding words as probability densities in a low-dimensional space. Rather than assuming that a word embedding is fixed across the entire text collection, as in standard word embedding methods, in our Bayesian model we generate it from a word-specific prior density for each occurrence of a gi...
Embedding Words as Distributions with a Bayesian Skip-gram Model
2,017
http://arxiv.org/pdf/1711.11027v1
Title Embedding Words Distributions Bayesian Skipgram Model Summary introduce method embedding word probability density lowdimensional space Rather assuming word embedding fixed across entire text collection standard word embedding method Bayesian model generate wordspecific prior density occurrence given word Intuitiv...
[0.002965732244774699, 0.06544508785009384, -0.008189809508621693, 0.04639604687690735, -0.02558886632323265, -0.0006541734328493476, -0.03012380562722683, -0.014476426877081394, -0.043928112834692, -0.02053113654255867, 0.008904443122446537, 0.004943918902426958, 0.03951584920287132, 0.05918828025460243, 0.04315352439...
1,672
1,672
['Li Zhou', 'Kevin Small', 'Oleg Rokhlenko', 'Charles Elkan']
1712.02838v1
Learning a goal-oriented dialog policy is generally performed offline with supervised learning algorithms or online with reinforcement learning (RL). Additionally, as companies accumulate massive quantities of dialog transcripts between customers and trained human agents, encoder-decoder methods have gained popularity ...
End-to-End Offline Goal-Oriented Dialog Policy Learning via Policy Gradient
2,017
http://arxiv.org/pdf/1712.02838v1
Title EndtoEnd Offline GoalOriented Dialog Policy Learning via Policy Gradient Summary Learning goaloriented dialog policy generally performed offline supervised learning algorithm online reinforcement learning RL Additionally company accumulate massive quantity dialog transcript customer trained human agent encoderdec...
[0.07026870548725128, 0.08006059378385544, 0.004132372792810202, -0.026854172348976135, -0.015680581331253052, 0.020831666886806488, 0.05506717786192894, -0.001994699938222766, 0.0016297514084726572, -0.04661821946501732, 0.010016540996730328, 0.0023116066586226225, -0.03365248069167137, 0.08696337044239044, -0.0247235...
1,673
1,673
['Tom Bocklisch', 'Joey Faulkner', 'Nick Pawlowski', 'Alan Nichol']
1712.05181v2
We introduce a pair of tools, Rasa NLU and Rasa Core, which are open source python libraries for building conversational software. Their purpose is to make machine-learning based dialogue management and language understanding accessible to non-specialist software developers. In terms of design philosophy, we aim for ea...
Rasa: Open Source Language Understanding and Dialogue Management
2,017
http://arxiv.org/pdf/1712.05181v2
Title Rasa Open Source Language Understanding Dialogue Management Summary introduce pair tool Rasa NLU Rasa Core open source python library building conversational software purpose make machinelearning based dialogue management language understanding accessible nonspecialist software developer term design philosophy ai...
[0.07375922799110413, 0.0342569425702095, -0.009282796643674374, 0.0536576583981514, -0.022660011425614357, 0.01786349155008793, 0.008081112056970596, -0.01835542544722557, -0.0007569087320007384, -0.05111638084053993, -0.014852745458483696, -0.012527520768344402, -0.0028458291199058294, 0.05537257716059685, -0.0318022...
1,674
1,674
['Jordan Prosky', 'Xingyou Song', 'Andrew Tan', 'Michael Zhao']
1712.05785v2
In this work, we present our findings and experiments for stock-market prediction using various textual sentiment analysis tools, such as mood analysis and event extraction, as well as prediction models, such as LSTMs and specific convolutional architectures.
Sentiment Predictability for Stocks
2,017
http://arxiv.org/pdf/1712.05785v2
Title Sentiment Predictability Stocks Summary work present finding experiment stockmarket prediction using various textual sentiment analysis tool mood analysis event extraction well prediction model LSTMs specific convolutional architecture Authors 0 Ahmed Osman Wojciech Samek 1 Ji Young Lee Franck Dernoncourt 2 Iulia...
[0.043276309967041016, 0.003283623605966568, -0.028049619868397713, 0.02579721063375473, 0.014050290919840336, -0.01893913932144642, -0.006835345644503832, 0.0338779017329216, 0.06133340671658516, -0.02586187794804573, -0.0005685030482709408, 0.005392367951571941, -0.05247477814555168, 0.06113961711525917, -0.010852052...
1,675
1,675
['Shaika Chowdhury', 'Chenwei Zhang', 'Philip S. Yu']
1801.06294v5
Social media has grown to be a crucial information source for pharmacovigilance studies where an increasing number of people post adverse reactions to medical drugs that are previously unreported. Aiming to effectively monitor various aspects of Adverse Drug Reactions (ADRs) from diversely expressed social medical post...
Multi-Task Pharmacovigilance Mining from Social Media Posts
2,018
http://arxiv.org/pdf/1801.06294v5
Title MultiTask Pharmacovigilance Mining Social Media Posts Summary Social medium grown crucial information source pharmacovigilance study increasing number people post adverse reaction medical drug previously unreported Aiming effectively monitor various aspect Adverse Drug Reactions ADRs diversely expressed social me...
[0.04507583752274513, 0.013317495584487915, -0.0074925716035068035, -0.057485658675432205, -0.001354943960905075, 0.03905411437153816, -0.006052253767848015, 0.017672574147582054, 0.04889145493507385, -0.017991188913583755, 0.021162470802664757, -0.02722763642668724, -0.05177292600274086, 0.05652874708175659, 0.0207503...
1,676
1,676
['Natalia Ruemmele', 'Yuriy Tyshetskiy', 'Alex Collins']
1801.09788v1
Relational data sources are still one of the most popular ways to store enterprise or Web data, however, the issue with relational schema is the lack of a well-defined semantic description. A common ontology provides a way to represent the meaning of a relational schema and can facilitate the integration of heterogeneo...
Evaluating approaches for supervised semantic labeling
2,018
http://arxiv.org/pdf/1801.09788v1
Title Evaluating approach supervised semantic labeling Summary Relational data source still one popular way store enterprise Web data however issue relational schema lack welldefined semantic description common ontology provides way represent meaning relational schema facilitate integration heterogeneous data source wi...
[0.03930589556694031, 0.015284717082977295, -0.010274292901158333, -0.010159802623093128, -0.004478284157812595, 0.017467588186264038, 0.015432214364409447, -0.008884986862540245, 0.0020922906696796417, -0.09602612257003784, -0.02278350107371807, 0.0460173636674881, -0.03393826633691788, 0.09379855543375015, -0.0173143...
1,677
1,677
['Xiaoyu Shen', 'Hui Su', 'Shuzi Niu', 'Vera Demberg']
1802.02032v1
Variational encoder-decoders (VEDs) have shown promising results in dialogue generation. However, the latent variable distributions are usually approximated by a much simpler model than the powerful RNN structure used for encoding and decoding, yielding the KL-vanishing problem and inconsistent training objective. In t...
Improving Variational Encoder-Decoders in Dialogue Generation
2,018
http://arxiv.org/pdf/1802.02032v1
Title Improving Variational EncoderDecoders Dialogue Generation Summary Variational encoderdecoders VEDs shown promising result dialogue generation However latent variable distribution usually approximated much simpler model powerful RNN structure used encoding decoding yielding KLvanishing problem inconsistent trainin...
[0.03610539808869362, 0.07636425644159317, 0.0009403272997587919, 0.027679147198796272, 0.023890579119324684, -0.005811105482280254, -0.012724149972200394, -0.025710204616189003, -0.03349072486162186, -0.02872893586754799, 0.010159006342291832, -0.04423771798610687, -0.00023354207223746926, 0.08470749855041504, 0.02481...
1,678
1,678
['Lajanugen Logeswaran', 'Honglak Lee']
1803.02893v1
In this work we propose a simple and efficient framework for learning sentence representations from unlabelled data. Drawing inspiration from the distributional hypothesis and recent work on learning sentence representations, we reformulate the problem of predicting the context in which a sentence appears as a classifi...
An efficient framework for learning sentence representations
2,018
http://arxiv.org/pdf/1803.02893v1
Title efficient framework learning sentence representation Summary work propose simple efficient framework learning sentence representation unlabelled data Drawing inspiration distributional hypothesis recent work learning sentence representation reformulate problem predicting context sentence appears classification pr...
[0.019252127036452293, 0.036825377494096756, 0.0001673534425208345, 0.08646602183580399, -0.0532311350107193, 0.016560088843107224, -0.011923360638320446, -0.010462495498359203, 0.00886057410389185, -0.11063437908887863, 0.003508961061015725, 0.01939280517399311, -0.017436755821108818, 0.023861950263381004, -0.01581116...
1,679
1,679
['Akira Taniguchi', 'Yoshinobu Hagiwara', 'Tadahiro Taniguchi', 'Tetsunari Inamura']
1803.03481v1
In this paper, we propose a novel online learning algorithm, SpCoSLAM 2.0 for spatial concepts and lexical acquisition with higher accuracy and scalability. In previous work, we proposed SpCoSLAM as an online learning algorithm based on the Rao--Blackwellized particle filter. However, this conventional algorithm had pr...
SpCoSLAM 2.0: An Improved and Scalable Online Learning of Spatial Concepts and Language Models with Mapping
2,018
http://arxiv.org/pdf/1803.03481v1
Title SpCoSLAM 20 Improved Scalable Online Learning Spatial Concepts Language Models Mapping Summary paper propose novel online learning algorithm SpCoSLAM 20 spatial concept lexical acquisition higher accuracy scalability previous work proposed SpCoSLAM online learning algorithm based RaoBlackwellized particle filter ...
[0.04308997467160225, -0.0034989984706044197, -0.005255053285509348, -0.01302257739007473, -0.003517488483339548, -0.005973104853183031, -0.010670684278011322, 0.008596866391599178, 0.005492329131811857, -0.05108483508229256, -0.011043794453144073, -0.01459698285907507, 0.06825695931911469, 0.053748637437820435, -0.005...
1,680
1,680
['Alon Talmor', 'Jonathan Berant']
1803.06643v1
Answering complex questions is a time-consuming activity for humans that requires reasoning and integration of information. Recent work on reading comprehension made headway in answering simple questions, but tackling complex questions is still an ongoing research challenge. Conversely, semantic parsers have been succe...
The Web as a Knowledge-base for Answering Complex Questions
2,018
http://arxiv.org/pdf/1803.06643v1
Title Web Knowledgebase Answering Complex Questions Summary Answering complex question timeconsuming activity human requires reasoning integration information Recent work reading comprehension made headway answering simple question tackling complex question still ongoing research challenge Conversely semantic parser su...
[0.06209965795278549, -0.0018522560130804777, -0.028064746409654617, 0.03913737088441849, -0.0010538778733462095, 0.036047156900167465, -0.012547679245471954, 0.011786721646785736, 0.0012314387131482363, -0.05244043841958046, 0.020607251673936844, -0.0077393255196511745, -0.0207138042896986, 0.024797918274998665, 0.000...
1,681
1,681
['Sidi Lu', 'Yaoming Zhu', 'Weinan Zhang', 'Jun Wang', 'Yong Yu']
1803.07133v1
This paper presents a systematic survey on recent development of neural text generation models. Specifically, we start from recurrent neural network language models with the traditional maximum likelihood estimation training scheme and point out its shortcoming for text generation. We thus introduce the recently propos...
Neural Text Generation: Past, Present and Beyond
2,018
http://arxiv.org/pdf/1803.07133v1
Title Neural Text Generation Past Present Beyond Summary paper present systematic survey recent development neural text generation model Specifically start recurrent neural network language model traditional maximum likelihood estimation training scheme point shortcoming text generation thus introduce recently proposed...
[0.03416601940989494, -0.0046037728898227215, 0.010822712443768978, 0.021766463294625282, 0.010045988485217094, -0.01917230896651745, 0.00803808681666851, 0.0031340993009507656, -0.007855450734496117, -0.018745236098766327, 0.019061004742980003, -0.04219358041882515, 0.0277880746871233, 0.029451211914420128, 0.02459779...
1,682
1,682
['Christopher M. White', 'Sanjeev P. Khudanpur', 'Patrick J. Wolfe']
0911.3944v1
In conventional supervised pattern recognition tasks, model selection is typically accomplished by minimizing the classification error rate on a set of so-called development data, subject to ground-truth labeling by human experts or some other means. In the context of speech processing systems and other large-scale pra...
Likelihood-based semi-supervised model selection with applications to speech processing
2,009
http://arxiv.org/pdf/0911.3944v1
Title Likelihoodbased semisupervised model selection application speech processing Summary conventional supervised pattern recognition task model selection typically accomplished minimizing classification error rate set socalled development data subject groundtruth labeling human expert mean context speech processing s...
[0.020036496222019196, 0.04514201730489731, 0.01891029067337513, 0.007164965849369764, -0.019941454753279686, 0.0039193895645439625, 0.07322937250137329, 0.007458778098225594, 0.022550711408257484, -0.015751516446471214, -0.0664927139878273, 0.028136158362030983, 0.0488026887178421, 0.03373991325497627, 0.0056360163725...
1,683
1,683
['Sebastian Riedel', 'David A. Smith', 'Andrew McCallum']
1203.3511v1
We speed up marginal inference by ignoring factors that do not significantly contribute to overall accuracy. In order to pick a suitable subset of factors to ignore, we propose three schemes: minimizing the number of model factors under a bound on the KL divergence between pruned and full models; minimizing the KL dive...
Inference by Minimizing Size, Divergence, or their Sum
2,012
http://arxiv.org/pdf/1203.3511v1
Title Inference Minimizing Size Divergence Sum Summary speed marginal inference ignoring factor significantly contribute overall accuracy order pick suitable subset factor ignore propose three scheme minimizing number model factor bound KL divergence pruned full model minimizing KL divergence bound factor count minimiz...
[0.0177320446819067, 0.020164459943771362, -0.004124578554183245, 0.024249807000160217, -0.023213855922222137, -0.0032474847976118326, 0.04022090137004852, -0.0009767285082489252, -0.03401361033320427, -0.007064837496727705, 0.0308056753128767, 0.026485230773687363, 0.05891337990760803, 0.033469703048467636, 0.00969137...
1,684
1,684
['Khalid El-Arini', 'Emily B. Fox', 'Carlos Guestrin']
1204.2523v1
In information retrieval, a fundamental goal is to transform a document into concepts that are representative of its content. The term "representative" is in itself challenging to define, and various tasks require different granularities of concepts. In this paper, we aim to model concepts that are sparse over the voca...
Concept Modeling with Superwords
2,012
http://arxiv.org/pdf/1204.2523v1
Title Concept Modeling Superwords Summary information retrieval fundamental goal transform document concept representative content term representative challenging define various task require different granularity concept paper aim model concept sparse vocabulary flexibly adapt content based relevant semantic informatio...
[0.06983456760644913, 0.0363796129822731, -0.004285464063286781, -0.015265509486198425, -0.0191810205578804, 0.0007976330234669149, 0.03161235898733139, 0.027442365884780884, -0.08558855205774307, -0.09219018369913101, 0.03553773835301399, -0.005280886311084032, -0.008684554137289524, 0.1005798801779747, -0.03289601579...
1,685
1,685
['Yanqing Chen', 'Bryan Perozzi', 'Rami Al-Rfou', 'Steven Skiena']
1301.3226v4
We seek to better understand the difference in quality of the several publicly released embeddings. We propose several tasks that help to distinguish the characteristics of different embeddings. Our evaluation of sentiment polarity and synonym/antonym relations shows that embeddings are able to capture surprisingly nua...
The Expressive Power of Word Embeddings
2,013
http://arxiv.org/pdf/1301.3226v4
Title Expressive Power Word Embeddings Summary seek better understand difference quality several publicly released embeddings propose several task help distinguish characteristic different embeddings evaluation sentiment polarity synonymantonym relation show embeddings able capture surprisingly nuanced semantics even a...
[0.015741199254989624, 0.024538015946745872, -0.011564748361706734, 0.04371295124292374, -0.036257002502679825, -0.00473531149327755, -0.06202245131134987, -0.022343970835208893, -0.041961558163166046, -0.048370689153671265, 0.004000484477728605, -0.0034569206181913614, 0.022754456847906113, 0.03134175390005112, 0.0341...
1,686
1,686
['Jeon-Hyung Kang', 'Jun Ma', 'Yan Liu']
1301.5686v2
The increasing volume of short texts generated on social media sites, such as Twitter or Facebook, creates a great demand for effective and efficient topic modeling approaches. While latent Dirichlet allocation (LDA) can be applied, it is not optimal due to its weakness in handling short texts with fast-changing topics...
Transfer Topic Modeling with Ease and Scalability
2,013
http://arxiv.org/pdf/1301.5686v2
Title Transfer Topic Modeling Ease Scalability Summary increasing volume short text generated social medium site Twitter Facebook creates great demand effective efficient topic modeling approach latent Dirichlet allocation LDA applied optimal due weakness handling short text fastchanging topic scalability concern paper...
[0.045756496489048004, 0.024798234924674034, -0.03031793050467968, 0.01262965053319931, -0.023088647052645683, 0.010709757916629314, 0.015666862949728966, -0.005654686596244574, -0.04006776213645935, -0.05918652564287186, -0.008307693526148796, 0.01734677515923977, -0.028061646968126297, 0.03882577270269394, -0.0245478...
1,687
1,687
['Hui Lin', 'Jeff A. Bilmes']
1210.4871v1
We introduce a method to learn a mixture of submodular "shells" in a large-margin setting. A submodular shell is an abstract submodular function that can be instantiated with a ground set and a set of parameters to produce a submodular function. A mixture of such shells can then also be so instantiated to produce a mor...
Learning Mixtures of Submodular Shells with Application to Document Summarization
2,012
http://arxiv.org/pdf/1210.4871v1
Title Learning Mixtures Submodular Shells Application Document Summarization Summary introduce method learn mixture submodular shell largemargin setting submodular shell abstract submodular function instantiated ground set set parameter produce submodular function mixture shell also instantiated produce complex submodu...
[0.03594502434134483, 0.026326436549425125, 0.005371913779526949, -0.01002875529229641, -0.01328013651072979, 0.014266877435147762, 0.0018255955073982477, 0.007665037643164396, -0.0439019501209259, -0.036801792681217194, 0.02063385397195816, 0.011812923476099968, 0.022489620372653008, 0.03894847631454468, 0.00145876815...
1,688
1,688
['Dani Yogatama', 'Manaal Faruqui', 'Chris Dyer', 'Noah A. Smith']
1406.2035v2
We propose a new method for learning word representations using hierarchical regularization in sparse coding inspired by the linguistic study of word meanings. We show an efficient learning algorithm based on stochastic proximal methods that is significantly faster than previous approaches, making it possible to perfor...
Learning Word Representations with Hierarchical Sparse Coding
2,014
http://arxiv.org/pdf/1406.2035v2
Title Learning Word Representations Hierarchical Sparse Coding Summary propose new method learning word representation using hierarchical regularization sparse coding inspired linguistic study word meaning show efficient learning algorithm based stochastic proximal method significantly faster previous approach making p...
[0.022380083799362183, 0.030501075088977814, 0.004449342377483845, 0.046074818819761276, -0.037925854325294495, 0.02717979997396469, -0.060099683701992035, 0.052897773683071136, -0.018428007140755653, -0.061853643506765366, -0.026108697056770325, -0.014350238256156445, 0.037642497569322586, 0.04682381451129913, 0.01324...
1,689
1,689
['Misha Denil', 'Alban Demiraj', 'Nal Kalchbrenner', 'Phil Blunsom', 'Nando de Freitas']
1406.3830v1
Capturing the compositional process which maps the meaning of words to that of documents is a central challenge for researchers in Natural Language Processing and Information Retrieval. We introduce a model that is able to represent the meaning of documents by embedding them in a low dimensional vector space, while pre...
Modelling, Visualising and Summarising Documents with a Single Convolutional Neural Network
2,014
http://arxiv.org/pdf/1406.3830v1
Title Modelling Visualising Summarising Documents Single Convolutional Neural Network Summary Capturing compositional process map meaning word document central challenge researcher Natural Language Processing Information Retrieval introduce model able represent meaning document embedding low dimensional vector space pr...
[0.04569587856531143, 0.022319355979561806, 0.002504092874005437, 0.04700756072998047, -0.016436459496617317, 0.007322603836655617, -0.009084511548280716, -0.007108236663043499, -0.038527294993400574, -0.002436610171571374, 0.012667061761021614, 0.03027123026549816, -0.012872609309852123, 0.06480460613965988, -0.014966...
1,690
1,690
['Santiago Segarra', 'Mark Eisen', 'Alejandro Ribeiro']
1406.4469v1
A method for authorship attribution based on function word adjacency networks (WANs) is introduced. Function words are parts of speech that express grammatical relationships between other words but do not carry lexical meaning on their own. In the WANs in this paper, nodes are function words and directed edges stand in...
Authorship Attribution through Function Word Adjacency Networks
2,014
http://arxiv.org/pdf/1406.4469v1
Title Authorship Attribution Function Word Adjacency Networks Summary method authorship attribution based function word adjacency network WANs introduced Function word part speech express grammatical relationship word carry lexical meaning WANs paper node function word directed edge stand likelihood finding sink word o...
[0.052150603383779526, 0.03523486480116844, -0.01252550445497036, 0.022812582552433014, -0.07045439630746841, -0.008146919310092926, 0.10468510538339615, -0.0003662681265268475, -0.016756195574998856, -0.017148952931165695, 0.003241403494030237, -0.0017514204373583198, 0.06309854984283447, 0.022164085879921913, 0.03672...
1,691
1,691
['Jordan Boyd-Graber', 'David Blei']
1205.2657v1
We develop the multilingual topic model for unaligned text (MuTo), a probabilistic model of text that is designed to analyze corpora composed of documents in two languages. From these documents, MuTo uses stochastic EM to simultaneously discover both a matching between the languages and multilingual latent topics. We d...
Multilingual Topic Models for Unaligned Text
2,012
http://arxiv.org/pdf/1205.2657v1
Title Multilingual Topic Models Unaligned Text Summary develop multilingual topic model unaligned text MuTo probabilistic model text designed analyze corpus composed document two language document MuTo us stochastic EM simultaneously discover matching language multilingual latent topic demonstrate MuTo able find shared...
[0.0401880219578743, 0.039453163743019104, -0.020931608974933624, 0.030335277318954468, -0.06315498799085617, 0.02395118959248066, 0.0329737514257431, -0.013979288749396801, -0.05989363044500351, -0.03411046415567398, -0.008619396016001701, -0.002986607374623418, 0.009932879358530045, 0.011730416677892208, -0.008662069...
1,692
1,692
['Pengtao Xie', 'Eric P. Xing']
1309.6874v1
Document clustering and topic modeling are two closely related tasks which can mutually benefit each other. Topic modeling can project documents into a topic space which facilitates effective document clustering. Cluster labels discovered by document clustering can be incorporated into topic models to extract local top...
Integrating Document Clustering and Topic Modeling
2,013
http://arxiv.org/pdf/1309.6874v1
Title Integrating Document Clustering Topic Modeling Summary Document clustering topic modeling two closely related task mutually benefit Topic modeling project document topic space facilitates effective document clustering Cluster label discovered document clustering incorporated topic model extract local topic specif...
[0.019635405391454697, 0.011050519533455372, -0.01615780033171177, -0.01147067453712225, -0.013106259517371655, 0.04417801648378372, 0.010947390459477901, -0.027285175397992134, -0.026134738698601723, -0.020673153921961784, -0.003828794229775667, 0.035643965005874634, 0.006979977246373892, 0.03287478908896446, 0.027982...
1,693
1,693
['Kayhan Batmanghelich', 'Ardavan Saeedi', 'Karthik Narasimhan', 'Sam Gershman']
1604.00126v1
Traditional topic models do not account for semantic regularities in language. Recent distributional representations of words exhibit semantic consistency over directional metrics such as cosine similarity. However, neither categorical nor Gaussian observational distributions used in existing topic models are appropria...
Nonparametric Spherical Topic Modeling with Word Embeddings
2,016
http://arxiv.org/pdf/1604.00126v1
Title Nonparametric Spherical Topic Modeling Word Embeddings Summary Traditional topic model account semantic regularity language Recent distributional representation word exhibit semantic consistency directional metric cosine similarity However neither categorical Gaussian observational distribution used existing topi...
[0.0057060145772993565, 0.022160755470395088, -0.0070287869311869144, 0.03228248283267021, -0.026390178129076958, -0.011601919308304787, -0.01534274686127901, -0.045775625854730606, -0.06940498948097229, -0.02348918840289116, 0.0006371333147399127, -0.016236066818237305, 0.03372505307197571, 0.03352741897106171, 0.0139...
1,694
1,694
['Ke Jiang', 'Suvrit Sra', 'Brian Kulis']
1604.02027v2
Topic models have emerged as fundamental tools in unsupervised machine learning. Most modern topic modeling algorithms take a probabilistic view and derive inference algorithms based on Latent Dirichlet Allocation (LDA) or its variants. In contrast, we study topic modeling as a combinatorial optimization problem, and p...
Combinatorial Topic Models using Small-Variance Asymptotics
2,016
http://arxiv.org/pdf/1604.02027v2
Title Combinatorial Topic Models using SmallVariance Asymptotics Summary Topic model emerged fundamental tool unsupervised machine learning modern topic modeling algorithm take probabilistic view derive inference algorithm based Latent Dirichlet Allocation LDA variant contrast study topic modeling combinatorial optimiz...
[0.024551568552851677, 0.01987987756729126, -0.03045199252665043, 0.04240508750081062, -0.01664101704955101, -0.012069480493664742, 0.01800502836704254, 0.006555367726832628, -0.05291632190346718, -0.04143618047237396, 0.012070857919752598, 0.011822933331131935, -0.00598755432292819, 0.007767262402921915, -0.0042690420...
1,695
1,695
['Nebojsa Jojic', 'Alessandro Perina']
1202.3752v1
Models of bags of words typically assume topic mixing so that the words in a single bag come from a limited number of topics. We show here that many sets of bag of words exhibit a very different pattern of variation than the patterns that are efficiently captured by topic mixing. In many cases, from one bag of words to...
Multidimensional counting grids: Inferring word order from disordered bags of words
2,012
http://arxiv.org/pdf/1202.3752v1
Title Multidimensional counting grid Inferring word order disordered bag word Summary Models bag word typically assume topic mixing word single bag come limited number topic show many set bag word exhibit different pattern variation pattern efficiently captured topic mixing many case one bag word next word disappear ne...
[0.02661585621535778, 0.06926091760396957, -0.02375229261815548, 0.0447484590113163, -0.011329682543873787, 0.003030281513929367, 0.0012402402935549617, 0.0012876333203166723, -0.04763896390795708, -0.018433721736073494, 0.03479677438735962, -0.030386749655008316, 0.008727922104299068, 0.0943383052945137, 0.00177532387...
1,696
1,696
['Amit Gruber', 'Michal Rosen-Zvi', 'Yair Weiss']
1206.3254v1
Latent topic models have been successfully applied as an unsupervised topic discovery technique in large document collections. With the proliferation of hypertext document collection such as the Internet, there has also been great interest in extending these approaches to hypertext [6, 9]. These approaches typically mo...
Latent Topic Models for Hypertext
2,012
http://arxiv.org/pdf/1206.3254v1
Title Latent Topic Models Hypertext Summary Latent topic model successfully applied unsupervised topic discovery technique large document collection proliferation hypertext document collection Internet also great interest extending approach hypertext 6 9 approach typically model link analogous fashion model word docume...
[0.05610544607043266, 0.007630609441548586, -0.04029342532157898, 0.018157267943024635, -0.022825250402092934, -0.0010487455874681473, -0.0031824675388634205, -0.004456654656678438, -0.015317131765186787, -0.04651668295264244, -0.007951859384775162, 0.029648592695593834, -0.023967014625668526, 0.0555642805993557, -0.00...
1,697
1,697
['Paul Prasse', 'Christoph Sawade', 'Niels Landwehr', 'Tobias Scheffer']
1206.4637v1
This paper addresses the problem of inferring a regular expression from a given set of strings that resembles, as closely as possible, the regular expression that a human expert would have written to identify the language. This is motivated by our goal of automating the task of postmasters of an email service who use r...
Learning to Identify Regular Expressions that Describe Email Campaigns
2,012
http://arxiv.org/pdf/1206.4637v1
Title Learning Identify Regular Expressions Describe Email Campaigns Summary paper address problem inferring regular expression given set string resembles closely possible regular expression human expert would written identify language motivated goal automating task postmaster email service use regular expression descr...
[0.06091081351041794, 0.07895389199256897, -0.02280392125248909, 0.03494830057024956, -0.05667182803153992, -0.004200463183224201, 0.03992246463894844, 0.016673384234309196, -0.011061637662351131, -0.08030781894922256, 0.04133298620581627, 0.009989321231842041, -0.040565818548202515, 0.13541382551193237, -0.02065250650...
1,698
1,698
['Stanislas Lauly', 'Alex Boulanger', 'Hugo Larochelle']
1401.1803v1
Recent work on learning multilingual word representations usually relies on the use of word-level alignements (e.g. infered with the help of GIZA++) between translated sentences, in order to align the word embeddings in different languages. In this workshop paper, we investigate an autoencoder model for learning multil...
Learning Multilingual Word Representations using a Bag-of-Words Autoencoder
2,014
http://arxiv.org/pdf/1401.1803v1
Title Learning Multilingual Word Representations using BagofWords Autoencoder Summary Recent work learning multilingual word representation usually relies use wordlevel alignements eg infered help GIZA translated sentence order align word embeddings different language workshop paper investigate autoencoder model learni...
[0.004452795255929232, 0.03080795146524906, -0.013707432895898819, 0.08919369429349899, -0.022904641926288605, 0.025647608563303947, 0.02996945008635521, -0.02703395113348961, -0.025189245119690895, -0.05778256803750992, -0.04741382226347923, -0.015068118460476398, 0.032449979335069656, 0.041521914303302765, 0.01344800...
1,699
1,699
['Hossein Soleimani', 'David J. Miller']
1401.6169v2
We propose a parsimonious topic model for text corpora. In related models such as Latent Dirichlet Allocation (LDA), all words are modeled topic-specifically, even though many words occur with similar frequencies across different topics. Our modeling determines salient words for each topic, which have topic-specific pr...
Parsimonious Topic Models with Salient Word Discovery
2,014
http://arxiv.org/pdf/1401.6169v2
Title Parsimonious Topic Models Salient Word Discovery Summary propose parsimonious topic model text corpus related model Latent Dirichlet Allocation LDA word modeled topicspecifically even though many word occur similar frequency across different topic modeling determines salient word topic topicspecific probability r...
[0.05725876986980438, 0.007782648783177137, 0.006330019794404507, 0.03112320601940155, -0.0548204630613327, 0.015298910439014435, 0.013084899634122849, 0.012756772339344025, -0.06528640538454056, -0.0505606010556221, -0.0018134000711143017, 0.03615998476743698, 0.007967738434672356, 0.025439713150262833, -0.00867627002...