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
1,500
1,500
['Kyung-Min Kim', 'Min-Oh Heo', 'Seong-Ho Choi', 'Byoung-Tak Zhang']
1707.00836v1
Question-answering (QA) on video contents is a significant challenge for achieving human-level intelligence as it involves both vision and language in real-world settings. Here we demonstrate the possibility of an AI agent performing video story QA by learning from a large amount of cartoon videos. We develop a video-s...
DeepStory: Video Story QA by Deep Embedded Memory Networks
2,017
http://arxiv.org/pdf/1707.00836v1
Title DeepStory Video Story QA Deep Embedded Memory Networks Summary Questionanswering QA video content significant challenge achieving humanlevel intelligence involves vision language realworld setting demonstrate possibility AI agent performing video story QA learning large amount cartoon video develop videostory lea...
[0.030133472755551338, 0.04644474759697914, 0.0059533980675041676, 0.06620924919843674, 0.009385606274008751, 0.019018415361642838, 0.01992093399167061, -0.007949190214276314, -0.05469324812293053, -0.04758176580071449, 0.02004476636648178, -0.027979036793112755, -0.006461616605520248, 0.09040877968072891, 0.0098303724...
1,501
1,501
['Emiel van Miltenburg', 'Desmond Elliott', 'Piek Vossen']
1707.01736v2
Automatic image description systems are commonly trained and evaluated on large image description datasets. Recently, researchers have started to collect such datasets for languages other than English. An unexplored question is how different these datasets are from English and, if there are any differences, what causes...
Cross-linguistic differences and similarities in image descriptions
2,017
http://arxiv.org/pdf/1707.01736v2
Title Crosslinguistic difference similarity image description Summary Automatic image description system commonly trained evaluated large image description datasets Recently researcher started collect datasets language English unexplored question different datasets English difference cause differ paper provides crossli...
[0.014726240187883377, 0.0323873907327652, -0.0022080608177930117, 0.055728670209646225, -0.03312385082244873, 0.03574204072356224, 0.04038747400045395, 0.0063042896799743176, -0.035782188177108765, -0.034163326025009155, -0.048508889973163605, -0.04353836923837662, 0.048604078590869904, -0.0012798893731087446, 0.01119...
1,502
1,502
['Prithvijit Chattopadhyay', 'Deshraj Yadav', 'Viraj Prabhu', 'Arjun Chandrasekaran', 'Abhishek Das', 'Stefan Lee', 'Dhruv Batra', 'Devi Parikh']
1708.05122v1
As AI continues to advance, human-AI teams are inevitable. However, progress in AI is routinely measured in isolation, without a human in the loop. It is crucial to benchmark progress in AI, not just in isolation, but also in terms of how it translates to helping humans perform certain tasks, i.e., the performance of h...
Evaluating Visual Conversational Agents via Cooperative Human-AI Games
2,017
http://arxiv.org/pdf/1708.05122v1
Title Evaluating Visual Conversational Agents via Cooperative HumanAI Games Summary AI continues advance humanAI team inevitable However progress AI routinely measured isolation without human loop crucial benchmark progress AI isolation also term translates helping human perform certain task ie performance humanAI team...
[0.04477175325155258, 0.03201380744576454, -0.04653344303369522, 0.029246579855680466, -0.03470310941338539, 0.014898104593157768, 0.08318676054477692, 0.021127087995409966, 0.012961364351212978, -0.006685969419777393, -0.03261445090174675, -0.010300660505890846, 0.04987237602472305, 0.07067786157131195, 0.005030433647...
1,503
1,503
['Lea Frermann', 'Shay B. Cohen', 'Mirella Lapata']
1710.11601v1
In this paper we argue that crime drama exemplified in television programs such as CSI:Crime Scene Investigation is an ideal testbed for approximating real-world natural language understanding and the complex inferences associated with it. We propose to treat crime drama as a new inference task, capitalizing on the fac...
Whodunnit? Crime Drama as a Case for Natural Language Understanding
2,017
http://arxiv.org/pdf/1710.11601v1
Title Whodunnit Crime Drama Case Natural Language Understanding Summary paper argue crime drama exemplified television program CSICrime Scene Investigation ideal testbed approximating realworld natural language understanding complex inference associated propose treat crime drama new inference task capitalizing fact epi...
[0.051760442554950714, 0.05733175575733185, -0.007958034984767437, 0.056677740067243576, -0.10429243743419647, 0.025099031627178192, 0.04071785882115364, 0.034174494445323944, -0.020982008427381516, -0.03295779600739479, 0.014337355270981789, 0.0005922110285609961, 0.014609538950026035, 0.0596160888671875, -0.021603336...
1,504
1,504
['Pan Lu', 'Hongsheng Li', 'Wei Zhang', 'Jianyong Wang', 'Xiaogang Wang']
1711.06794v2
Recently, the Visual Question Answering (VQA) task has gained increasing attention in artificial intelligence. Existing VQA methods mainly adopt the visual attention mechanism to associate the input question with corresponding image regions for effective question answering. The free-form region based and the detection-...
Co-attending Free-form Regions and Detections with Multi-modal Multiplicative Feature Embedding for Visual Question Answering
2,017
http://arxiv.org/pdf/1711.06794v2
Title Coattending Freeform Regions Detections Multimodal Multiplicative Feature Embedding Visual Question Answering Summary Recently Visual Question Answering VQA task gained increasing attention artificial intelligence Existing VQA method mainly adopt visual attention mechanism associate input question corresponding i...
[0.0005170387448742986, 0.0012836400419473648, -0.005757738836109638, 0.060082003474235535, -0.0003570384287741035, 0.010179027915000916, 0.015444103628396988, -0.0033848995808511972, 0.017354190349578857, -0.01706869900226593, 0.008239361457526684, -0.037031419575214386, -0.020928850397467613, 0.07414920628070831, 0.0...
1,505
1,505
['Peter Anderson', 'Qi Wu', 'Damien Teney', 'Jake Bruce', 'Mark Johnson', 'Niko Sünderhauf', 'Ian Reid', 'Stephen Gould', 'Anton van den Hengel']
1711.07280v2
A robot that can carry out a natural-language instruction has been a dream since before the Jetsons cartoon series imagined a life of leisure mediated by a fleet of attentive robot helpers. It is a dream that remains stubbornly distant. However, recent advances in vision and language methods have made incredible progre...
Vision-and-Language Navigation: Interpreting visually-grounded navigation instructions in real environments
2,017
http://arxiv.org/pdf/1711.07280v2
Title VisionandLanguage Navigation Interpreting visuallygrounded navigation instruction real environment Summary robot carry naturallanguage instruction dream since Jetsons cartoon series imagined life leisure mediated fleet attentive robot helper dream remains stubbornly distant However recent advance vision language ...
[0.046394359320402145, 0.006258328910917044, 0.0020021668169647455, 0.026700399816036224, -0.016509249806404114, 0.04056892916560173, 0.006515354849398136, -0.019153006374835968, -0.030934149399399757, -0.05569329485297203, -0.025087321177124977, -0.0010000671027228236, 0.05233997106552124, 0.05448349565267563, 0.02178...
1,506
1,506
['Qi Wu', 'Peng Wang', 'Chunhua Shen', 'Ian Reid', 'Anton van den Hengel']
1711.07613v1
The Visual Dialogue task requires an agent to engage in a conversation about an image with a human. It represents an extension of the Visual Question Answering task in that the agent needs to answer a question about an image, but it needs to do so in light of the previous dialogue that has taken place. The key challeng...
Are You Talking to Me? Reasoned Visual Dialog Generation through Adversarial Learning
2,017
http://arxiv.org/pdf/1711.07613v1
Title Talking Reasoned Visual Dialog Generation Adversarial Learning Summary Visual Dialogue task requires agent engage conversation image human represents extension Visual Question Answering task agent need answer question image need light previous dialogue taken place key challenge Visual Dialogue thus maintaining co...
[0.05427750200033188, 0.08763076364994049, -0.012250179424881935, 0.013629692606627941, 0.0025183509569615126, -0.006586266681551933, 0.01352756842970848, 0.00927368551492691, -0.013711264356970787, -0.007238287478685379, 0.0018960463348776102, -0.0010001035407185555, -0.014145895838737488, 0.0625782385468483, 0.057231...
1,507
1,507
['Zhiqian Chen', 'Xuchao Zhang', 'Arnold P. Boedihardjo', 'Jing Dai', 'Chang-Tien Lu']
1712.01455v1
Deriving event storylines is an effective summarization method to succinctly organize extensive information, which can significantly alleviate the pain of information overload. The critical challenge is the lack of widely recognized definition of storyline metric. Prior studies have developed various approaches based o...
Multimodal Storytelling via Generative Adversarial Imitation Learning
2,017
http://arxiv.org/pdf/1712.01455v1
Title Multimodal Storytelling via Generative Adversarial Imitation Learning Summary Deriving event storyline effective summarization method succinctly organize extensive information significantly alleviate pain information overload critical challenge lack widely recognized definition storyline metric Prior study develo...
[0.04049281030893326, 0.04573928564786911, 0.0019740338902920485, -0.005928633268922567, -0.00984366238117218, -0.005590612534433603, 0.02581949159502983, 0.005309034138917923, -0.008802425116300583, -0.019213788211345673, -0.013643124140799046, -0.010985864326357841, -0.002810086589306593, 0.08799909055233002, 0.02718...
1,508
1,508
['Alexander Trott', 'Caiming Xiong', 'Richard Socher']
1712.08697v2
Questions that require counting a variety of objects in images remain a major challenge in visual question answering (VQA). The most common approaches to VQA involve either classifying answers based on fixed length representations of both the image and question or summing fractional counts estimated from each section o...
Interpretable Counting for Visual Question Answering
2,017
http://arxiv.org/pdf/1712.08697v2
Title Interpretable Counting Visual Question Answering Summary Questions require counting variety object image remain major challenge visual question answering VQA common approach VQA involve either classifying answer based fixed length representation image question summing fractional count estimated section image cont...
[0.03543373569846153, 0.04563778638839722, -0.025540655478835106, 0.0726831778883934, 0.0005311763961799443, -0.008116568438708782, 0.008720716461539268, 0.029185324907302856, -0.027653254568576813, 0.01625850796699524, 0.03498781472444534, -0.03390749171376228, -0.0022171048913151026, 0.08103693276643753, 0.0029668768...
1,509
1,509
['D. Kiela', 'E. Grave', 'A. Joulin', 'T. Mikolov']
1802.02892v1
While the incipient internet was largely text-based, the modern digital world is becoming increasingly multi-modal. Here, we examine multi-modal classification where one modality is discrete, e.g. text, and the other is continuous, e.g. visual representations transferred from a convolutional neural network. In particul...
Efficient Large-Scale Multi-Modal Classification
2,018
http://arxiv.org/pdf/1802.02892v1
Title Efficient LargeScale MultiModal Classification Summary incipient internet largely textbased modern digital world becoming increasingly multimodal examine multimodal classification one modality discrete eg text continuous eg visual representation transferred convolutional neural network particular focus scenario a...
[0.015183696523308754, 0.03291900455951691, 0.019272953271865845, 0.045037757605314255, -0.012969346717000008, 0.010724516585469246, 0.03774917125701904, 0.03590969368815422, -0.003371232422068715, -0.11698762327432632, -0.01875237375497818, -0.026959272101521492, 0.004661093465983868, 0.05317787826061249, 0.0067250058...
1,510
1,510
['Dong Huk Park', 'Lisa Anne Hendricks', 'Zeynep Akata', 'Anna Rohrbach', 'Bernt Schiele', 'Trevor Darrell', 'Marcus Rohrbach']
1802.08129v1
Deep models that are both effective and explainable are desirable in many settings; prior explainable models have been unimodal, offering either image-based visualization of attention weights or text-based generation of post-hoc justifications. We propose a multimodal approach to explanation, and argue that the two mod...
Multimodal Explanations: Justifying Decisions and Pointing to the Evidence
2,018
http://arxiv.org/pdf/1802.08129v1
Title Multimodal Explanations Justifying Decisions Pointing Evidence Summary Deep model effective explainable desirable many setting prior explainable model unimodal offering either imagebased visualization attention weight textbased generation posthoc justification propose multimodal approach explanation argue two mod...
[0.014485442079603672, 0.05350092798471451, -0.023613663390278816, 0.010680987499654293, -0.00047452186117880046, 0.028066810220479965, 0.07112065702676773, 0.006929676979780197, -0.06205486133694649, -0.0259800236672163, 0.03182191774249077, 0.015040453523397446, 0.011555949226021767, 0.03273480013012886, 0.0357989557...
1,511
1,511
['Albert Gatt', 'Marc Tanti', 'Adrian Muscat', 'Patrizia Paggio', 'Reuben A. Farrugia', 'Claudia Borg', 'Kenneth P. Camilleri', 'Mike Rosner', 'Lonneke van der Plas']
1803.03827v1
The past few years have witnessed renewed interest in NLP tasks at the interface between vision and language. One intensively-studied problem is that of automatically generating text from images. In this paper, we extend this problem to the more specific domain of face description. Unlike scene descriptions, face descr...
Face2Text: Collecting an Annotated Image Description Corpus for the Generation of Rich Face Descriptions
2,018
http://arxiv.org/pdf/1803.03827v1
Title Face2Text Collecting Annotated Image Description Corpus Generation Rich Face Descriptions Summary past year witnessed renewed interest NLP task interface vision language One intensivelystudied problem automatically generating text image paper extend problem specific domain face description Unlike scene descriptio...
[0.03659816086292267, 0.044442713260650635, 0.014160356484353542, 0.04890827462077141, -0.002589929848909378, 0.050650015473365784, 0.014903363771736622, 0.006570793222635984, -0.01331760548055172, -0.07147450000047684, 0.022033287212252617, -0.02245093509554863, 0.044645655900239944, 0.07403817772865295, 0.02483928017...
1,512
1,512
['Jasdeep Singh', 'Vincent Ying', 'Alex Nutkiewicz']
1803.07724v1
Visual Question Answering (VQA) is an increasingly popular topic in deep learning research, requiring coordination of natural language processing and computer vision modules into a single architecture. We build upon the model which placed first in the VQA Challenge by developing thirteen new attention mechanisms and in...
Attention on Attention: Architectures for Visual Question Answering (VQA)
2,018
http://arxiv.org/pdf/1803.07724v1
Title Attention Attention Architectures Visual Question Answering VQA Summary Visual Question Answering VQA increasingly popular topic deep learning research requiring coordination natural language processing computer vision module single architecture build upon model placed first VQA Challenge developing thirteen new ...
[0.07112914323806763, 0.013132777996361256, -0.032669659703969955, 0.02875717543065548, 6.635287718381733e-05, 0.011726823635399342, 0.016010846942663193, 0.009898804128170013, 0.010879400186240673, -0.025566691532731056, -0.010386679321527481, -0.022353388369083405, 0.00180678756441921, 0.03823559358716011, 0.05654412...
1,513
1,513
['Xin Wang', 'Wenhan Xiong', 'Hongmin Wang', 'William Yang Wang']
1803.07729v1
Existing research studies on vision and language grounding for robot navigation focus on improving model-free deep reinforcement learning (DRL) models in synthetic environments. However, model-free DRL models do not consider the dynamics in the real-world environments, and they often fail to generalize to new scenes. I...
Look Before You Leap: Bridging Model-Free and Model-Based Reinforcement Learning for Planned-Ahead Vision-and-Language Navigation
2,018
http://arxiv.org/pdf/1803.07729v1
Title Look Leap Bridging ModelFree ModelBased Reinforcement Learning PlannedAhead VisionandLanguage Navigation Summary Existing research study vision language grounding robot navigation focus improving modelfree deep reinforcement learning DRL model synthetic environment However modelfree DRL model consider dynamic rea...
[0.021762551739811897, 0.03991891071200371, 0.015614230185747147, -0.02709508314728737, -0.02253391034901142, 0.026505114510655403, 0.006790190935134888, -0.0079093798995018, -0.005695043597370386, -0.02005928009748459, -0.04458358883857727, 0.010105646215379238, -0.0100356824696064, 0.10479284077882767, 0.010328652337...
1,514
1,514
['N. Siddharth', 'Andrei Barbu', 'Jeffrey Mark Siskind']
1308.4189v2
We present a system that demonstrates how the compositional structure of events, in concert with the compositional structure of language, can interplay with the underlying focusing mechanisms in video action recognition, thereby providing a medium, not only for top-down and bottom-up integration, but also for multi-mod...
Seeing What You're Told: Sentence-Guided Activity Recognition In Video
2,013
http://arxiv.org/pdf/1308.4189v2
Title Seeing Youre Told SentenceGuided Activity Recognition Video Summary present system demonstrates compositional structure event concert compositional structure language interplay underlying focusing mechanism video action recognition thereby providing medium topdown bottomup integration also multimodal integration ...
[0.026433298364281654, -0.01934034377336502, -0.010754087008535862, 0.053804151713848114, -0.0015838064718991518, -0.004726769402623177, 0.029460253193974495, 0.0016095697646960616, -0.04277385398745537, -0.10204672068357468, -0.014101682230830193, -0.017604373395442963, 0.03253701701760292, 0.07194077968597412, 0.0274...
1,515
1,515
['Ting-Hao', 'Huang', 'Francis Ferraro', 'Nasrin Mostafazadeh', 'Ishan Misra', 'Aishwarya Agrawal', 'Jacob Devlin', 'Ross Girshick', 'Xiaodong He', 'Pushmeet Kohli', 'Dhruv Batra', 'C. Lawrence Zitnick', 'Devi Parikh', 'Lucy Vanderwende', 'Michel Galley', 'Margaret Mitchell']
1604.03968v1
We introduce the first dataset for sequential vision-to-language, and explore how this data may be used for the task of visual storytelling. The first release of this dataset, SIND v.1, includes 81,743 unique photos in 20,211 sequences, aligned to both descriptive (caption) and story language. We establish several stro...
Visual Storytelling
2,016
http://arxiv.org/pdf/1604.03968v1
Title Visual Storytelling Summary introduce first dataset sequential visiontolanguage explore data may used task visual storytelling first release dataset SIND v1 includes 81743 unique photo 20211 sequence aligned descriptive caption story language establish several strong baseline storytelling task motivate automatic ...
[0.031262923032045364, 0.045357510447502136, -0.004578910768032074, 0.01689263805747032, -0.011498530395328999, 0.03554327413439751, 0.01069460529834032, -0.016540002077817917, -0.031002923846244812, -0.03039037622511387, 0.046662814915180206, 0.003582028904929757, 0.04849691689014435, 0.09578617662191391, -0.016469454...
1,516
1,516
['Darko Brodic', 'Alessia Amelio', 'Zoran N. Milivojevic']
1509.01978v1
The paper presents a new script classification method for the discrimination of the South Slavic medieval labels. It consists in the textural analysis of the script types. In the first step, each letter is coded by the equivalent script type, which is defined by its typographical features. Obtained coded text is subjec...
An Approach to the Analysis of the South Slavic Medieval Labels Using Image Texture
2,015
http://arxiv.org/pdf/1509.01978v1
Title Approach Analysis South Slavic Medieval Labels Using Image Texture Summary paper present new script classification method discrimination South Slavic medieval label consists textural analysis script type first step letter coded equivalent script type defined typographical feature Obtained coded text subjected run...
[0.03283483162522316, -0.0007733716629445553, -0.012245564721524715, 0.09061595052480698, -0.038986023515462875, 0.014659491367638111, 0.04343695193529129, 0.0600128248333931, 0.03893284499645233, -0.024317581206560135, 0.07487364113330841, 0.011879277415573597, 0.05168822035193443, -0.03003777377307415, 0.030877953395...
1,517
1,517
['David Harwath', 'James Glass']
1511.03690v1
In this paper, we present a model which takes as input a corpus of images with relevant spoken captions and finds a correspondence between the two modalities. We employ a pair of convolutional neural networks to model visual objects and speech signals at the word level, and tie the networks together with an embedding a...
Deep Multimodal Semantic Embeddings for Speech and Images
2,015
http://arxiv.org/pdf/1511.03690v1
Title Deep Multimodal Semantic Embeddings Speech Images Summary paper present model take input corpus image relevant spoken caption find correspondence two modality employ pair convolutional neural network model visual object speech signal word level tie network together embedding alignment model learns joint semantic ...
[-0.00873226672410965, 0.07543084770441055, 0.015057925134897232, 0.06607715785503387, -0.010831444524228573, -0.011773885227739811, 0.0013816257705911994, -0.022349882870912552, -0.05505695566534996, -0.06997966766357422, -0.03555690869688988, 0.02152065746486187, 0.019578034058213234, 0.038855016231536865, 0.03835095...
1,518
1,518
['Douwe Kiela', 'Luana Bulat', 'Anita L. Vero', 'Stephen Clark']
1610.07432v1
Meaning has been called the "holy grail" of a variety of scientific disciplines, ranging from linguistics to philosophy, psychology and the neurosciences. The field of Artifical Intelligence (AI) is very much a part of that list: the development of sophisticated natural language semantics is a sine qua non for achievin...
Virtual Embodiment: A Scalable Long-Term Strategy for Artificial Intelligence Research
2,016
http://arxiv.org/pdf/1610.07432v1
Title Virtual Embodiment Scalable LongTerm Strategy Artificial Intelligence Research Summary Meaning called holy grail variety scientific discipline ranging linguistics philosophy psychology neuroscience field Artifical Intelligence AI much part list development sophisticated natural language semantics sine qua non ach...
[0.03190549463033676, 0.05104323476552963, -0.03199467435479164, -0.004489298909902573, -0.018653709441423416, -0.0005416027270257473, 0.05439896881580353, 0.025877615436911583, -0.01686997525393963, -0.038482144474983215, 0.04499237239360809, -0.019349021837115288, 0.02375146746635437, 0.036173265427351, 0.04851341620...
1,519
1,519
['Hideki Nakayama', 'Noriki Nishida']
1611.04503v3
We propose an approach to build a neural machine translation system with no supervised resources (i.e., no parallel corpora) using multimodal embedded representation over texts and images. Based on the assumption that text documents are often likely to be described with other multimedia information (e.g., images) somew...
Zero-resource Machine Translation by Multimodal Encoder-decoder Network with Multimedia Pivot
2,016
http://arxiv.org/pdf/1611.04503v3
Title Zeroresource Machine Translation Multimodal Encoderdecoder Network Multimedia Pivot Summary propose approach build neural machine translation system supervised resource ie parallel corpus using multimodal embedded representation text image Based assumption text document often likely described multimedia informati...
[0.014936608262360096, 0.010619210079312325, 0.012717947363853455, 0.07209783047437668, -0.03723142668604851, 0.0290552768856287, 0.016013137996196747, 0.027914321050047874, -0.06497275084257126, -0.06266327202320099, -0.033165983855724335, 0.01745942048728466, 0.016929255798459053, 0.008869967423379421, 0.042575057595...
1,520
1,520
['Satwik Kottur', 'José M. F. Moura', 'Stefan Lee', 'Dhruv Batra']
1706.08502v3
A number of recent works have proposed techniques for end-to-end learning of communication protocols among cooperative multi-agent populations, and have simultaneously found the emergence of grounded human-interpretable language in the protocols developed by the agents, all learned without any human supervision! In t...
Natural Language Does Not Emerge 'Naturally' in Multi-Agent Dialog
2,017
http://arxiv.org/pdf/1706.08502v3
Title Natural Language Emerge Naturally MultiAgent Dialog Summary number recent work proposed technique endtoend learning communication protocol among cooperative multiagent population simultaneously found emergence grounded humaninterpretable language protocol developed agent learned without human supervision paper us...
[0.06871876865625381, 0.03477110341191292, -0.03175624459981918, 0.006926555186510086, -0.025550760328769684, 0.002000841312110424, -0.015382152982056141, 0.0009209050913341343, -0.005151013843715191, -0.02861676551401615, -0.02094591222703457, 0.022132666781544685, 0.012186368927359581, 0.06348131597042084, 0.03508357...
1,521
1,521
['Yunan Ye', 'Zhou Zhao', 'Yimeng Li', 'Long Chen', 'Jun Xiao', 'Yueting Zhuang']
1707.06355v1
Video Question Answering is a challenging problem in visual information retrieval, which provides the answer to the referenced video content according to the question. However, the existing visual question answering approaches mainly tackle the problem of static image question, which may be ineffectively for video ques...
Video Question Answering via Attribute-Augmented Attention Network Learning
2,017
http://arxiv.org/pdf/1707.06355v1
Title Video Question Answering via AttributeAugmented Attention Network Learning Summary Video Question Answering challenging problem visual information retrieval provides answer referenced video content according question However existing visual question answering approach mainly tackle problem static image question m...
[0.0012210751883685589, -0.003758843056857586, -0.0193500816822052, 0.037479281425476074, 0.022408191114664078, -0.002933765761554241, 0.03415048494935036, 0.002558613196015358, -0.029608124867081642, -0.03784941881895065, 0.005661449860781431, -0.01162508875131607, -0.022577956318855286, 0.034490544348955154, 0.041863...
1,522
1,522
['Éloi Zablocki', 'Benjamin Piwowarski', 'Laure Soulier', 'Patrick Gallinari']
1711.03483v1
Representing the semantics of words is a long-standing problem for the natural language processing community. Most methods compute word semantics given their textual context in large corpora. More recently, researchers attempted to integrate perceptual and visual features. Most of these works consider the visual appear...
Learning Multi-Modal Word Representation Grounded in Visual Context
2,017
http://arxiv.org/pdf/1711.03483v1
Title Learning MultiModal Word Representation Grounded Visual Context Summary Representing semantics word longstanding problem natural language processing community method compute word semantics given textual context large corpus recently researcher attempted integrate perceptual visual feature work consider visual app...
[0.00917705800384283, 0.0021182845812290907, 0.011063353158533573, 0.08448777347803116, -0.02934405952692032, -0.007013994734734297, -0.003196458565071225, 0.02104269340634346, -0.018866874277591705, -0.09629105776548386, -0.01660628244280815, 0.016951395198702812, 0.01683768816292286, 0.061108000576496124, 0.039709448...
1,523
1,523
['Ying Hua Tan', 'Chee Seng Chan']
1711.05557v1
Automatic generation of caption to describe the content of an image has been gaining a lot of research interests recently, where most of the existing works treat the image caption as pure sequential data. Natural language, however possess a temporal hierarchy structure, with complex dependencies between each subsequenc...
Phrase-based Image Captioning with Hierarchical LSTM Model
2,017
http://arxiv.org/pdf/1711.05557v1
Title Phrasebased Image Captioning Hierarchical LSTM Model Summary Automatic generation caption describe content image gaining lot research interest recently existing work treat image caption pure sequential data Natural language however posse temporal hierarchy structure complex dependency subsequence paper propose ph...
[0.05603024363517761, 0.06615810841321945, -0.002028798218816519, 0.05084509402513504, -0.03974999114871025, 0.022061850875616074, 0.004179257433861494, 0.002193397842347622, -0.03064454346895218, -0.06547734141349792, -0.002777515910565853, -0.015096023678779602, 0.018236223608255386, 0.07573260366916656, 0.0092136980...
1,524
1,524
['Junjie Zhang', 'Qi Wu', 'Chunhua Shen', 'Jian Zhang', 'Jianfeng Lu', 'Anton van den Hengel']
1711.07614v1
Despite significant progress in a variety of vision-and-language problems, developing a method capable of asking intelligent, goal-oriented questions about images is proven to be an inscrutable challenge. Towards this end, we propose a Deep Reinforcement Learning framework based on three new intermediate rewards, namel...
Asking the Difficult Questions: Goal-Oriented Visual Question Generation via Intermediate Rewards
2,017
http://arxiv.org/pdf/1711.07614v1
Title Asking Difficult Questions GoalOriented Visual Question Generation via Intermediate Rewards Summary Despite significant progress variety visionandlanguage problem developing method capable asking intelligent goaloriented question image proven inscrutable challenge Towards end propose Deep Reinforcement Learning f...
[0.05706040561199188, 0.053884562104940414, -0.011793949641287327, 0.00976491253823042, -0.00841380376368761, 0.016033843159675598, 0.014405874535441399, 0.008488531224429607, -0.03280009329319, 0.026193544268608093, 0.009293919429183006, -0.0057765343226492405, -0.027176329866051674, 0.07293970882892609, 0.02995842695...
1,525
1,525
['Xin Wang', 'Wenhu Chen', 'Jiawei Wu', 'Yuan-Fang Wang', 'William Yang Wang']
1711.11135v2
Video captioning is the task of automatically generating a textual description of the actions in a video. Although previous work (e.g. sequence-to-sequence model) has shown promising results in abstracting a coarse description of a short video, it is still very challenging to caption a video containing multiple fine-gr...
Video Captioning via Hierarchical Reinforcement Learning
2,017
http://arxiv.org/pdf/1711.11135v2
Title Video Captioning via Hierarchical Reinforcement Learning Summary Video captioning task automatically generating textual description action video Although previous work eg sequencetosequence model shown promising result abstracting coarse description short video still challenging caption video containing multiple ...
[0.05518227815628052, 0.02546113356947899, -0.0066934432834386826, 0.012222967110574245, 0.0032720433082431555, 0.03646324947476387, -0.0014765849336981773, -0.01409946195781231, -0.06197468936443329, -0.0515042282640934, -0.013944936916232109, -0.021596871316432953, 0.0013697179965674877, 0.05844774842262268, -0.00717...
1,526
1,526
['Licheng Yu', 'Zhe Lin', 'Xiaohui Shen', 'Jimei Yang', 'Xin Lu', 'Mohit Bansal', 'Tamara L. Berg']
1801.08186v2
In this paper, we address referring expression comprehension: localizing an image region described by a natural language expression. While most recent work treats expressions as a single unit, we propose to decompose them into three modular components related to subject appearance, location, and relationship to other o...
MAttNet: Modular Attention Network for Referring Expression Comprehension
2,018
http://arxiv.org/pdf/1801.08186v2
Title MAttNet Modular Attention Network Referring Expression Comprehension Summary paper address referring expression comprehension localizing image region described natural language expression recent work treat expression single unit propose decompose three modular component related subject appearance location relatio...
[0.050801604986190796, 0.013055199757218361, 0.009616962634027004, 0.024465957656502724, 0.0004428372485563159, -0.01647518016397953, 0.024678681045770645, -0.006058621220290661, -0.057257767766714096, -0.04774269461631775, -0.0451434887945652, 0.047264017164707184, 0.007083653938025236, 0.07564333826303482, 0.04255723...
1,527
1,527
['Gangeshwar Krishnamurthy', 'Navonil Majumder', 'Soujanya Poria', 'Erik Cambria']
1803.00344v1
Automatic deception detection is an important task that has gained momentum in computational linguistics due to its potential applications. In this paper, we propose a simple yet tough to beat multi-modal neural model for deception detection. By combining features from different modalities such as video, audio, and tex...
A Deep Learning Approach for Multimodal Deception Detection
2,018
http://arxiv.org/pdf/1803.00344v1
Title Deep Learning Approach Multimodal Deception Detection Summary Automatic deception detection important task gained momentum computational linguistics due potential application paper propose simple yet tough beat multimodal neural model deception detection combining feature different modality video audio text along...
[0.0471406914293766, 0.06244907155632973, -0.04592951759696007, 0.09848599135875702, -0.059334151446819305, 0.007372592575848103, 0.040104709565639496, 0.026616627350449562, -0.026575881987810135, -0.03638509660959244, -0.022302808240056038, -0.010618518106639385, 0.04825775325298309, 0.055357009172439575, 0.0454686507...
1,528
1,528
['Li Yao', 'Nicolas Ballas', 'Kyunghyun Cho', 'John R. Smith', 'Yoshua Bengio']
1511.04590v5
The task of associating images and videos with a natural language description has attracted a great amount of attention recently. Rapid progress has been made in terms of both developing novel algorithms and releasing new datasets. Indeed, the state-of-the-art results on some of the standard datasets have been pushed i...
Oracle performance for visual captioning
2,015
http://arxiv.org/pdf/1511.04590v5
Title Oracle performance visual captioning Summary task associating image video natural language description attracted great amount attention recently Rapid progress made term developing novel algorithm releasing new datasets Indeed stateoftheart result standard datasets pushed regime become difficult make significant ...
[0.051785457879304886, 0.025116179138422012, 0.003318097908049822, 0.053381603211164474, -0.008077227510511875, 0.024572012946009636, -0.02039930783212185, 0.02165072411298752, -0.024584205821156502, -0.09302832186222076, -0.0176702793687582, -0.0034613902680575848, 0.03406914323568344, 0.06662707030773163, 0.006752479...
1,529
1,529
['Jason Poulos', 'Rafael Valle']
1712.04046v1
The paper approaches the problem of image-to-text with attention-based encoder-decoder networks that are trained to handle sequences of characters rather than words. We experiment on lines of text from a popular handwriting database with different attention mechanisms for the decoder. The model trained with softmax att...
Attention networks for image-to-text
2,017
http://arxiv.org/pdf/1712.04046v1
Title Attention network imagetotext Summary paper approach problem imagetotext attentionbased encoderdecoder network trained handle sequence character rather word experiment line text popular handwriting database different attention mechanism decoder model trained softmax attention achieves lowest test error outperform...
[0.0019007285591214895, 0.05617862939834595, 0.03453725203871727, 0.06279102712869644, 0.006475950125604868, 0.011985531076788902, 0.030356530100107193, 0.014710545539855957, -0.017746536061167717, 0.00512318592518568, -0.013929417356848717, -0.04242204502224922, 0.03129710257053375, 0.040295831859111786, 0.03620330989...
1,530
1,530
['Serguei A. Mokhov', 'Stephen Sinclair', 'Ian Clément', 'Dimitrios Nicolacopoulos', 'for the MARF R&D Group']
0905.1235v2
MARF is an open-source research platform and a collection of voice/sound/speech/text and natural language processing (NLP) algorithms written in Java and arranged into a modular and extensible framework facilitating addition of new algorithms. MARF can run distributively over the network and may act as a library in app...
The Modular Audio Recognition Framework (MARF) and its Applications: Scientific and Software Engineering Notes
2,009
http://arxiv.org/pdf/0905.1235v2
Title Modular Audio Recognition Framework MARF Applications Scientific Software Engineering Notes Summary MARF opensource research platform collection voicesoundspeechtext natural language processing NLP algorithm written Java arranged modular extensible framework facilitating addition new algorithm MARF run distributi...
[0.021967390552163124, 0.0023215003311634064, -0.00395343080163002, 0.03376096859574318, -0.020528629422187805, -0.007148134522140026, 0.023963794112205505, 0.008147230371832848, -0.025886524468660355, -0.010714133270084858, -0.043319087475538254, 0.06392878293991089, 0.04497349634766579, 0.02685609459877014, -0.008815...
1,531
1,531
['Mats Rooth', 'Stefan Riezler', 'Detlef Prescher', 'Glenn Carroll', 'Franz Beil']
cs/9905008v1
We present a technique for automatic induction of slot annotations for subcategorization frames, based on induction of hidden classes in the EM framework of statistical estimation. The models are empirically evalutated by a general decision test. Induction of slot labeling for subcategorization frames is accomplished b...
Inducing a Semantically Annotated Lexicon via EM-Based Clustering
1,999
http://arxiv.org/pdf/cs/9905008v1
Title Inducing Semantically Annotated Lexicon via EMBased Clustering Summary present technique automatic induction slot annotation subcategorization frame based induction hidden class EM framework statistical estimation model empirically evalutated general decision test Induction slot labeling subcategorization frame a...
[0.05431952700018883, -0.0049303253181278706, -0.011486358940601349, 0.038881562650203705, 0.0018617736641317606, 0.05658656358718872, -0.009410994127392769, -0.020894501358270645, 0.018234502524137497, -0.06522545963525772, 0.017605047672986984, 0.052051376551389694, -0.022141799330711365, -0.004953331314027309, 0.022...
1,532
1,532
['Yair Even-Zohar', 'Dan Roth']
cs/0009027v1
The eventual goal of a language model is to accurately predict the value of a missing word given its context. We present an approach to word prediction that is based on learning a representation for each word as a function of words and linguistics predicates in its context. This approach raises a few new questions that...
A Classification Approach to Word Prediction
2,000
http://arxiv.org/pdf/cs/0009027v1
Title Classification Approach Word Prediction Summary eventual goal language model accurately predict value missing word given context present approach word prediction based learning representation word function word linguistics predicate context approach raise new question address First order learn good word represent...
[0.06630980223417282, 0.03533277288079262, -0.013481303118169308, 0.037649523466825485, -0.009449375793337822, -0.019195619970560074, -0.0019921334460377693, 0.03718253970146179, 0.004436692222952843, -0.049508895725011826, -0.0041274260729551315, -0.010334674268960953, 0.022061536088585854, 0.05292689800262451, 0.0037...
1,533
1,533
['Yair Even-Zohar', 'Dan Roth']
cs/0106044v1
Many classification problems require decisions among a large number of competing classes. These tasks, however, are not handled well by general purpose learning methods and are usually addressed in an ad-hoc fashion. We suggest a general approach -- a sequential learning model that utilizes classifiers to sequentially ...
A Sequential Model for Multi-Class Classification
2,001
http://arxiv.org/pdf/cs/0106044v1
Title Sequential Model MultiClass Classification Summary Many classification problem require decision among large number competing class task however handled well general purpose learning method usually addressed adhoc fashion suggest general approach sequential learning model utilizes classifier sequentially restrict ...
[-0.009281281381845474, 0.05377946421504021, -0.02298317477107048, -0.01377402525395155, -0.026861362159252167, -0.029946645721793175, 0.03882648050785065, 0.010047642514109612, -0.008037589490413666, -0.03710797056555748, 0.034605782479047775, -0.012825149111449718, -0.036644097417593, -0.0018106657080352306, -0.04123...
1,534
1,534
['Ruben Sipos', 'Pannaga Shivaswamy', 'Thorsten Joachims']
1110.2162v2
In this paper, we present a supervised learning approach to training submodular scoring functions for extractive multi-document summarization. By taking a structured predicition approach, we provide a large-margin method that directly optimizes a convex relaxation of the desired performance measure. The learning method...
Large-Margin Learning of Submodular Summarization Methods
2,011
http://arxiv.org/pdf/1110.2162v2
Title LargeMargin Learning Submodular Summarization Methods Summary paper present supervised learning approach training submodular scoring function extractive multidocument summarization taking structured predicition approach provide largemargin method directly optimizes convex relaxation desired performance measure le...
[0.002559960586950183, 0.019929882138967514, 0.01378955040127039, 0.019688734784722328, -0.0028620886150747538, 0.008074906654655933, -0.005112542305141687, 0.029508275911211967, -0.016941700130701065, -0.053454961627721786, 0.02075539529323578, 0.011725310236215591, 0.007783990819007158, 0.007029594853520393, -0.00287...
1,535
1,535
['Peter D. Turney']
1309.4035v1
Given appropriate representations of the semantic relations between carpenter and wood and between mason and stone (for example, vectors in a vector space model), a suitable algorithm should be able to recognize that these relations are highly similar (carpenter is to wood as mason is to stone; the relations are analog...
Domain and Function: A Dual-Space Model of Semantic Relations and Compositions
2,013
http://arxiv.org/pdf/1309.4035v1
Title Domain Function DualSpace Model Semantic Relations Compositions Summary Given appropriate representation semantic relation carpenter wood mason stone example vector vector space model suitable algorithm able recognize relation highly similar carpenter wood mason stone relation analogous Likewise representation do...
[-0.0059359073638916016, 0.04831380397081375, -0.034959711134433746, 0.05660812556743622, -0.039942365139722824, 0.01838095858693123, 0.0089960228651762, 0.022512491792440414, -0.034252796322107315, -0.045976925641298294, -0.01155065931379795, 0.04097605124115944, -0.006540137808769941, 0.005689786747097969, -0.0099486...
1,536
1,536
['Mo Yu', 'Mark Dredze', 'Raman Arora', 'Matthew Gormley']
1604.00461v1
Modern NLP models rely heavily on engineered features, which often combine word and contextual information into complex lexical features. Such combination results in large numbers of features, which can lead to over-fitting. We present a new model that represents complex lexical features---comprised of parts for words,...
Embedding Lexical Features via Low-Rank Tensors
2,016
http://arxiv.org/pdf/1604.00461v1
Title Embedding Lexical Features via LowRank Tensors Summary Modern NLP model rely heavily engineered feature often combine word contextual information complex lexical feature combination result large number feature lead overfitting present new model represents complex lexical featurescomprised part word contextual inf...
[0.04098699986934662, 0.012918002903461456, 0.004055155906826258, 0.06386125087738037, -0.013983039185404778, 0.013886856846511364, 0.00031403257162310183, 0.015358430333435535, 0.04331746697425842, -0.06810712814331055, -0.0011290161637589335, -0.010153337381780148, 0.025406792759895325, -0.0036101657897233963, 0.0112...
1,537
1,537
['Rahul Kamath', 'Masanao Ochi', 'Yutaka Matsuo']
1604.05468v1
Online user reviews describing various products and services are now abundant on the web. While the information conveyed through review texts and ratings is easily comprehensible, there is a wealth of hidden information in them that is not immediately obvious. In this study, we unlock this hidden value behind user revi...
Understanding Rating Behaviour and Predicting Ratings by Identifying Representative Users
2,016
http://arxiv.org/pdf/1604.05468v1
Title Understanding Rating Behaviour Predicting Ratings Identifying Representative Users Summary Online user review describing various product service abundant web information conveyed review text rating easily comprehensible wealth hidden information immediately obvious study unlock hidden value behind user review und...
[0.04383717477321625, 0.03461770340800285, -0.03306559845805168, -0.0026462797541171312, -0.022476179525256157, 0.007777544669806957, 0.009503893554210663, 0.03514435142278671, -0.027214640751481056, -0.06690715998411179, -0.04867009073495865, -0.01669701375067234, 0.006711106281727552, 0.07849331200122833, -0.00292895...
1,538
1,538
['S. R. K. Branavan', 'David Silver', 'Regina Barzilay']
1401.5390v1
Domain knowledge is crucial for effective performance in autonomous control systems. Typically, human effort is required to encode this knowledge into a control algorithm. In this paper, we present an approach to language grounding which automatically interprets text in the context of a complex control application, suc...
Learning to Win by Reading Manuals in a Monte-Carlo Framework
2,014
http://arxiv.org/pdf/1401.5390v1
Title Learning Win Reading Manuals MonteCarlo Framework Summary Domain knowledge crucial effective performance autonomous control system Typically human effort required encode knowledge control algorithm paper present approach language grounding automatically interprets text context complex control application game us ...
[0.03243567422032356, 0.01777425780892372, -0.04072120040655136, 0.06688117235898972, -0.010690166614949703, -0.009499150328338146, -0.030728351324796677, 0.01206942182034254, -0.027517762035131454, -0.027609243988990784, -0.00378386746160686, 0.01764768175780773, 0.00975474901497364, 0.07678088545799255, 0.02632593177...
1,539
1,539
['Peter D. Turney', 'Saif M. Mohammad']
1401.8269v1
Inference in natural language often involves recognizing lexical entailment (RLE); that is, identifying whether one word entails another. For example, "buy" entails "own". Two general strategies for RLE have been proposed: One strategy is to manually construct an asymmetric similarity measure for context vectors (direc...
Experiments with Three Approaches to Recognizing Lexical Entailment
2,014
http://arxiv.org/pdf/1401.8269v1
Title Experiments Three Approaches Recognizing Lexical Entailment Summary Inference natural language often involves recognizing lexical entailment RLE identifying whether one word entail another example buy entail Two general strategy RLE proposed One strategy manually construct asymmetric similarity measure context ve...
[0.04271933063864708, 0.008541004732251167, -0.010344751179218292, 0.09168219566345215, -0.06608854979276657, 0.024685226380825043, -0.043492887169122696, 0.04584610089659691, 0.0021394945215433836, -0.04644320160150528, 0.009998542256653309, 0.04862793907523155, 0.039300091564655304, -0.010721194557845592, -0.02195796...
1,540
1,540
['Phong Le', 'Willem Zuidema']
1503.02510v2
We are proposing an extension of the recursive neural network that makes use of a variant of the long short-term memory architecture. The extension allows information low in parse trees to be stored in a memory register (the `memory cell') and used much later higher up in the parse tree. This provides a solution to the...
Compositional Distributional Semantics with Long Short Term Memory
2,015
http://arxiv.org/pdf/1503.02510v2
Title Compositional Distributional Semantics Long Short Term Memory Summary proposing extension recursive neural network make use variant long shortterm memory architecture extension allows information low parse tree stored memory register memory cell used much later higher parse tree provides solution vanishing gradie...
[0.0187027957290411, 0.05861368030309677, 0.009186853654682636, 0.044407859444618225, -0.023538833484053612, -0.004024796653538942, -0.05861158296465874, -0.03708852455019951, -0.03923381492495537, -0.06118423491716385, 0.0020083209965378046, -0.033779576420784, 0.018974188715219498, 0.028962712734937668, -0.0070997523...
1,541
1,541
['I. Lopez-Gazpio', 'M. Maritxalar', 'A. Gonzalez-Agirre', 'G. Rigau', 'L. Uria', 'E. Agirre']
1612.04868v1
User acceptance of artificial intelligence agents might depend on their ability to explain their reasoning, which requires adding an interpretability layer that fa- cilitates users to understand their behavior. This paper focuses on adding an in- terpretable layer on top of Semantic Textual Similarity (STS), which meas...
Interpretable Semantic Textual Similarity: Finding and explaining differences between sentences
2,016
http://arxiv.org/pdf/1612.04868v1
Title Interpretable Semantic Textual Similarity Finding explaining difference sentence Summary User acceptance artificial intelligence agent might depend ability explain reasoning requires adding interpretability layer fa cilitates user understand behavior paper focus adding terpretable layer top Semantic Textual Simil...
[0.04790806397795677, 0.024043599143624306, -0.031796906143426895, 0.0566096231341362, -0.04913182184100151, 0.005525683984160423, 0.015540756285190582, -0.00046548101818189025, -0.023098714649677277, -0.026324523612856865, 0.00966890063136816, 0.05728849396109581, -0.00024313069297932088, 0.04802684485912323, -0.00689...
1,542
1,542
['Xiujun Li', 'Zachary C. Lipton', 'Bhuwan Dhingra', 'Lihong Li', 'Jianfeng Gao', 'Yun-Nung Chen']
1612.05688v3
Despite widespread interests in reinforcement-learning for task-oriented dialogue systems, several obstacles can frustrate research and development progress. First, reinforcement learners typically require interaction with the environment, so conventional dialogue corpora cannot be used directly. Second, each task pres...
A User Simulator for Task-Completion Dialogues
2,016
http://arxiv.org/pdf/1612.05688v3
Title User Simulator TaskCompletion Dialogues Summary Despite widespread interest reinforcementlearning taskoriented dialogue system several obstacle frustrate research development progress First reinforcement learner typically require interaction environment conventional dialogue corpus cannot used directly Second tas...
[0.08492454886436462, 0.019643837586045265, -0.02458796836435795, 0.016287904232740402, -0.021420910954475403, 0.00900997407734394, 0.025990858674049377, -0.01344034168869257, -0.009310419671237469, -0.037980977445840836, -0.054618678987026215, -0.01609417423605919, -0.011970088817179203, 0.05443693324923515, -0.000441...
1,543
1,543
['Linas Vepstas']
1310.6775v1
This report describes the suicidality prediction models created under the DARPA DCAPS program in association with the Durkheim Project [http://durkheimproject.org/]. The models were built primarily from unstructured text (free-format clinician notes) for several hundred patient records obtained from the Veterans Health...
Durkheim Project Data Analysis Report
2,013
http://arxiv.org/pdf/1310.6775v1
Title Durkheim Project Data Analysis Report Summary report describes suicidality prediction model created DARPA DCAPS program association Durkheim Project httpdurkheimprojectorg model built primarily unstructured text freeformat clinician note several hundred patient record obtained Veterans Health Administration VHA m...
[0.053112391382455826, 0.04058682546019554, 0.007839888334274292, -0.05879482254385948, -0.00370315695181489, 0.05413030460476875, -0.0049728951416909695, 0.018170583993196487, 0.05849464237689972, 0.011485894210636616, 0.0377836748957634, 0.010038142092525959, 0.001560777542181313, 0.027801692485809326, -0.05484363064...
1,544
1,544
['Alberto Garcia-Duran', 'Antoine Bordes', 'Nicolas Usunier', 'Yves Grandvalet']
1506.00999v1
This paper tackles the problem of endogenous link prediction for Knowledge Base completion. Knowledge Bases can be represented as directed graphs whose nodes correspond to entities and edges to relationships. Previous attempts either consist of powerful systems with high capacity to model complex connectivity patterns,...
Combining Two And Three-Way Embeddings Models for Link Prediction in Knowledge Bases
2,015
http://arxiv.org/pdf/1506.00999v1
Title Combining Two ThreeWay Embeddings Models Link Prediction Knowledge Bases Summary paper tackle problem endogenous link prediction Knowledge Base completion Knowledge Bases represented directed graph whose node correspond entity edge relationship Previous attempt either consist powerful system high capacity model c...
[0.028028691187500954, 0.012016288004815578, -0.007711803074926138, 0.041000090539455414, 0.02017020992934704, -0.01689310185611248, -0.026607876643538475, 0.008887025527656078, 0.08558936417102814, -0.019767358899116516, 0.0021875451784580946, 0.06725065410137177, -0.038270849734544754, 0.03147250786423683, 0.05813724...
1,545
1,545
['Andrew M. Dai', 'Christopher Olah', 'Quoc V. Le']
1507.07998v1
Paragraph Vectors has been recently proposed as an unsupervised method for learning distributed representations for pieces of texts. In their work, the authors showed that the method can learn an embedding of movie review texts which can be leveraged for sentiment analysis. That proof of concept, while encouraging, was...
Document Embedding with Paragraph Vectors
2,015
http://arxiv.org/pdf/1507.07998v1
Title Document Embedding Paragraph Vectors Summary Paragraph Vectors recently proposed unsupervised method learning distributed representation piece text work author showed method learn embedding movie review text leveraged sentiment analysis proof concept encouraging rather narrow consider task sentiment analysis prov...
[0.02966325916349888, 0.024960463866591454, 0.015457241795957088, 0.06362469494342804, -0.02618514746427536, 0.03405144065618515, -0.0004784599004779011, 0.0015235008904710412, -0.006360932718962431, -0.05719435587525368, -0.003349226899445057, 0.01718319021165371, -0.006742198020219803, -0.017576249316334724, 0.018777...
1,546
1,546
['Sylvester Olubolu Orimaye', 'Saadat M. Alhashmi', 'Eu-Gene Siew', 'Sang Jung Kang']
1511.02385v1
We propose an effective technique to solving review-level sentiment classification problem by using sentence-level polarity correction. Our polarity correction technique takes into account the consistency of the polarities (positive and negative) of sentences within each product review before performing the actual mach...
Review-Level Sentiment Classification with Sentence-Level Polarity Correction
2,015
http://arxiv.org/pdf/1511.02385v1
Title ReviewLevel Sentiment Classification SentenceLevel Polarity Correction Summary propose effective technique solving reviewlevel sentiment classification problem using sentencelevel polarity correction polarity correction technique take account consistency polarity positive negative sentence within product review p...
[0.05685720965266228, 0.02480805292725563, -0.017736373469233513, 0.0037765074521303177, -0.07624901831150055, 0.011770646087825298, -0.01676351949572563, -0.002707916544750333, -0.04103584587574005, -0.10011200606822968, -0.011045793071389198, 0.02097293734550476, 0.03001459129154682, 0.051093779504299164, -0.06282084...
1,547
1,547
['Mrinmaya Sachan', 'Avinava Dubey', 'Eric P. Xing']
1602.04375v2
We provide a solution for elementary science test using instructional materials. We posit that there is a hidden structure that explains the correctness of an answer given the question and instructional materials and present a unified max-margin framework that learns to find these hidden structures (given a corpus of q...
Science Question Answering using Instructional Materials
2,016
http://arxiv.org/pdf/1602.04375v2
Title Science Question Answering using Instructional Materials Summary provide solution elementary science test using instructional material posit hidden structure explains correctness answer given question instructional material present unified maxmargin framework learns find hidden structure given corpus questionansw...
[0.056278277188539505, -0.022029511630535126, -0.011030483059585094, 0.02812047116458416, -0.017327843233942986, 0.0064567746594548225, 0.0380975641310215, 7.063688826747239e-05, -0.005643642973154783, -0.034807268530130386, 0.02868095599114895, 0.010268548503518105, 0.0055223689414560795, -0.018972786143422127, 0.0295...
1,548
1,548
['Franck Dernoncourt', 'Ji Young Lee', 'Trung H. Bui', 'Hung H. Bui']
1605.02130v1
The Dialog State Tracking Challenge 4 (DSTC 4) differentiates itself from the previous three editions as follows: the number of slot-value pairs present in the ontology is much larger, no spoken language understanding output is given, and utterances are labeled at the subdialog level. This paper describes a novel dialo...
Robust Dialog State Tracking for Large Ontologies
2,016
http://arxiv.org/pdf/1605.02130v1
Title Robust Dialog State Tracking Large Ontologies Summary Dialog State Tracking Challenge 4 DSTC 4 differentiates previous three edition follows number slotvalue pair present ontology much larger spoken language understanding output given utterance labeled subdialog level paper describes novel dialog state tracking m...
[0.03826859965920448, 0.06366908550262451, -0.012861158698797226, 0.04239136353135109, -0.030235031619668007, -0.03593390807509422, -0.01756763644516468, 0.04151202738285065, 0.02271481230854988, -0.07830679416656494, -0.002512074774131179, -0.0068204873241484165, -0.014039174653589725, 0.08903896808624268, -0.04596791...
1,549
1,549
['Jason D. Williams', 'Geoffrey Zweig']
1606.01269v1
This paper presents a model for end-to-end learning of task-oriented dialog systems. The main component of the model is a recurrent neural network (an LSTM), which maps from raw dialog history directly to a distribution over system actions. The LSTM automatically infers a representation of dialog history, which relieve...
End-to-end LSTM-based dialog control optimized with supervised and reinforcement learning
2,016
http://arxiv.org/pdf/1606.01269v1
Title Endtoend LSTMbased dialog control optimized supervised reinforcement learning Summary paper present model endtoend learning taskoriented dialog system main component model recurrent neural network LSTM map raw dialog history directly distribution system action LSTM automatically infers representation dialog histo...
[0.051641665399074554, 0.01784227229654789, -0.004451015032827854, 0.008939031511545181, -0.020290512591600418, 0.006012679077684879, 0.02966233901679516, -0.01972254179418087, 0.011466304771602154, -0.04880748689174652, -0.03614175692200661, -0.017550090327858925, -0.03669380396604538, 0.10485812276601791, -0.01345213...
1,550
1,550
['Gemma Boleda', 'Sebastian Padó', 'Marco Baroni']
1606.08777v1
One of the most basic functions of language is to refer to objects in a shared scene. Modeling reference with continuous representations is challenging because it requires individuation, i.e., tracking and distinguishing an arbitrary number of referents. We introduce a neural network model that, given a definite descri...
"Show me the cup": Reference with Continuous Representations
2,016
http://arxiv.org/pdf/1606.08777v1
Title Show cup Reference Continuous Representations Summary One basic function language refer object shared scene Modeling reference continuous representation challenging requires individuation ie tracking distinguishing arbitrary number referent introduce neural network model given definite description set object repr...
[-0.006277420092374086, 0.02291712909936905, -0.0038892508018761873, 0.0548073910176754, -0.03327573463320732, -0.007200880907475948, 0.03877125680446625, 0.008315987884998322, -0.038652170449495316, -0.05001630261540413, 0.029751285910606384, 0.0120850233361125, 0.023664837703108788, 0.09595131129026413, 0.01775722764...
1,551
1,551
['Yusuke Watanabe', 'Kazuma Hashimoto', 'Yoshimasa Tsuruoka']
1607.00410v1
We propose a simple domain adaptation method for neural networks in a supervised setting. Supervised domain adaptation is a way of improving the generalization performance on the target domain by using the source domain dataset, assuming that both of the datasets are labeled. Recently, recurrent neural networks have be...
Domain Adaptation for Neural Networks by Parameter Augmentation
2,016
http://arxiv.org/pdf/1607.00410v1
Title Domain Adaptation Neural Networks Parameter Augmentation Summary propose simple domain adaptation method neural network supervised setting Supervised domain adaptation way improving generalization performance target domain using source domain dataset assuming datasets labeled Recently recurrent neural network sho...
[0.04336482286453247, -0.0041063399985432625, -0.011841155588626862, 0.029627079144120216, 0.020963573828339577, -0.01588670164346695, 0.01790582202374935, -0.02461034245789051, -0.029105262830853462, -0.05511215329170227, -0.06734620034694672, -0.008666292764246464, 0.0026396862231194973, 0.05739433318376541, 0.007248...
1,552
1,552
['Hao Cheng', 'Hao Fang', 'Xiaodong He', 'Jianfeng Gao', 'Li Deng']
1608.02076v2
We develop a novel bi-directional attention model for dependency parsing, which learns to agree on headword predictions from the forward and backward parsing directions. The parsing procedure for each direction is formulated as sequentially querying the memory component that stores continuous headword embeddings. The p...
Bi-directional Attention with Agreement for Dependency Parsing
2,016
http://arxiv.org/pdf/1608.02076v2
Title Bidirectional Attention Agreement Dependency Parsing Summary develop novel bidirectional attention model dependency parsing learns agree headword prediction forward backward parsing direction parsing procedure direction formulated sequentially querying memory component store continuous headword embeddings propose...
[0.060248248279094696, 0.04773513972759247, -0.0028152584563940763, 0.04942653700709343, -0.037927694618701935, -0.0072561693377792835, -0.009004265069961548, -0.045214828103780746, 0.03480297327041626, -0.06175129488110542, -0.0006681988597847521, -0.016256099566817284, -0.015917887911200523, 0.03934372588992119, 0.00...
1,553
1,553
['Yonghui Wu', 'Mike Schuster', 'Zhifeng Chen', 'Quoc V. Le', 'Mohammad Norouzi', 'Wolfgang Macherey', 'Maxim Krikun', 'Yuan Cao', 'Qin Gao', 'Klaus Macherey', 'Jeff Klingner', 'Apurva Shah', 'Melvin Johnson', 'Xiaobing Liu', 'Łukasz Kaiser', 'Stephan Gouws', 'Yoshikiyo Kato', 'Taku Kudo', 'Hideto Kazawa', 'Keith Steve...
1609.08144v2
Neural Machine Translation (NMT) is an end-to-end learning approach for automated translation, with the potential to overcome many of the weaknesses of conventional phrase-based translation systems. Unfortunately, NMT systems are known to be computationally expensive both in training and in translation inference. Also,...
Google's Neural Machine Translation System: Bridging the Gap between Human and Machine Translation
2,016
http://arxiv.org/pdf/1609.08144v2
Title Googles Neural Machine Translation System Bridging Gap Human Machine Translation Summary Neural Machine Translation NMT endtoend learning approach automated translation potential overcome many weakness conventional phrasebased translation system Unfortunately NMT system known computationally expensive training tr...
[0.05189327150583267, 0.02330116741359234, -0.008303668349981308, 0.04859665036201477, -0.03353251516819, 0.009001576341688633, 0.02375219576060772, 0.022388502955436707, -0.022952023893594742, -0.024727340787649155, 0.0215552169829607, -0.02546580322086811, 0.020117294043302536, 0.04865515977144241, 0.0549696795642375...
1,554
1,554
['Andrea F. Daniele', 'Mohit Bansal', 'Matthew R. Walter']
1610.03164v1
Modern robotics applications that involve human-robot interaction require robots to be able to communicate with humans seamlessly and effectively. Natural language provides a flexible and efficient medium through which robots can exchange information with their human partners. Significant advancements have been made in...
Navigational Instruction Generation as Inverse Reinforcement Learning with Neural Machine Translation
2,016
http://arxiv.org/pdf/1610.03164v1
Title Navigational Instruction Generation Inverse Reinforcement Learning Neural Machine Translation Summary Modern robotics application involve humanrobot interaction require robot able communicate human seamlessly effectively Natural language provides flexible efficient medium robot exchange information human partner ...
[0.079897940158844, -0.017648454755544662, 0.002054268028587103, -0.045762013643980026, -0.031182950362563133, -0.0034264090936630964, -0.008073522709310055, -0.014271177351474762, 0.024013886228203773, -0.021959956735372543, -0.02507353574037552, -0.002367755863815546, 0.031878288835287094, 0.04860537871718407, 0.0318...
1,555
1,555
['Pranay Dighe', 'Afsaneh Asaei', 'Herve Bourlard']
1610.05688v1
Conventional deep neural networks (DNN) for speech acoustic modeling rely on Gaussian mixture models (GMM) and hidden Markov model (HMM) to obtain binary class labels as the targets for DNN training. Subword classes in speech recognition systems correspond to context-dependent tied states or senones. The present work a...
Low-rank and Sparse Soft Targets to Learn Better DNN Acoustic Models
2,016
http://arxiv.org/pdf/1610.05688v1
Title Lowrank Sparse Soft Targets Learn Better DNN Acoustic Models Summary Conventional deep neural network DNN speech acoustic modeling rely Gaussian mixture model GMM hidden Markov model HMM obtain binary class label target DNN training Subword class speech recognition system correspond contextdependent tied state se...
[-0.014875013381242752, 0.07498429715633392, 0.016875483095645905, 0.03213956952095032, 0.018765462562441826, -0.019199058413505554, 0.05762359872460365, -0.0063454643823206425, -0.08877308666706085, -0.016316192224621773, -0.045490339398384094, 0.0010008546523749828, 0.05620955303311348, 0.03414314612746239, 0.0317935...
1,556
1,556
['Emilio Jorge', 'Mikael Kågebäck', 'Fredrik D. Johansson', 'Emil Gustavsson']
1611.03218v4
Acquiring your first language is an incredible feat and not easily duplicated. Learning to communicate using nothing but a few pictureless books, a corpus, would likely be impossible even for humans. Nevertheless, this is the dominating approach in most natural language processing today. As an alternative, we propose t...
Learning to Play Guess Who? and Inventing a Grounded Language as a Consequence
2,016
http://arxiv.org/pdf/1611.03218v4
Title Learning Play Guess Inventing Grounded Language Consequence Summary Acquiring first language incredible feat easily duplicated Learning communicate using nothing pictureless book corpus would likely impossible even human Nevertheless dominating approach natural language processing today alternative propose use si...
[0.06512670964002609, 0.010973848402500153, -0.01911110058426857, 0.0338449664413929, -0.02581559680402279, -0.013170906342566013, 0.026614341884851456, 0.003021910088136792, -0.021852826699614525, -0.04921087622642517, -0.010073494166135788, 0.025341274216771126, 0.012789956294000149, 0.07466502487659454, 0.0212932657...
1,557
1,557
['Tianran Hu', 'Haoyuan Xiao', 'Thuy-vy Thi Nguyen', 'Jiebo Luo']
1701.06233v1
Many aspects of people's lives are proven to be deeply connected to their jobs. In this paper, we first investigate the distinct characteristics of major occupation categories based on tweets. From multiple social media platforms, we gather several types of user information. From users' LinkedIn webpages, we learn thei...
What the Language You Tweet Says About Your Occupation
2,017
http://arxiv.org/pdf/1701.06233v1
Title Language Tweet Says Occupation Summary Many aspect people life proven deeply connected job paper first investigate distinct characteristic major occupation category based tweet multiple social medium platform gather several type user information user LinkedIn webpage learn proficiency overcome ambiguity selfrepor...
[0.04466227442026138, 0.06082230806350708, -0.02260933257639408, -0.004650297574698925, 0.007393775507807732, 0.016344252973794937, 0.030874237418174744, -0.002101134741678834, -0.0031162407249212265, -0.04572472348809242, 0.03440006077289581, -0.00414568604901433, 0.00537402369081974, 0.03944060206413269, -0.001005281...
1,558
1,558
['Upol Ehsan', 'Brent Harrison', 'Larry Chan', 'Mark O. Riedl']
1702.07826v2
We introduce AI rationalization, an approach for generating explanations of autonomous system behavior as if a human had performed the behavior. We describe a rationalization technique that uses neural machine translation to translate internal state-action representations of an autonomous agent into natural language. W...
Rationalization: A Neural Machine Translation Approach to Generating Natural Language Explanations
2,017
http://arxiv.org/pdf/1702.07826v2
Title Rationalization Neural Machine Translation Approach Generating Natural Language Explanations Summary introduce AI rationalization approach generating explanation autonomous system behavior human performed behavior describe rationalization technique us neural machine translation translate internal stateaction repr...
[0.012970753014087677, 0.012885996140539646, -0.024664554744958878, 0.03993220999836922, -0.004605880472809076, -0.01348373293876648, -0.0027028885670006275, 0.02589661441743374, -0.0506434328854084, -0.01639520563185215, 0.03432780131697655, 0.030868399888277054, 0.046406183391809464, 0.048184268176555634, 0.042769115...
1,559
1,559
['Guokun Lai', 'Qizhe Xie', 'Hanxiao Liu', 'Yiming Yang', 'Eduard Hovy']
1704.04683v5
We present RACE, a new dataset for benchmark evaluation of methods in the reading comprehension task. Collected from the English exams for middle and high school Chinese students in the age range between 12 to 18, RACE consists of near 28,000 passages and near 100,000 questions generated by human experts (English instr...
RACE: Large-scale ReAding Comprehension Dataset From Examinations
2,017
http://arxiv.org/pdf/1704.04683v5
Title RACE Largescale ReAding Comprehension Dataset Examinations Summary present RACE new dataset benchmark evaluation method reading comprehension task Collected English exam middle high school Chinese student age range 12 18 RACE consists near 28000 passage near 100000 question generated human expert English instruct...
[0.003057262394577265, -0.005577275529503822, -0.027806589379906654, 0.033995553851127625, -0.006342193577438593, 0.046297796070575714, 0.05402746424078941, -0.006598419509828091, 0.002527610631659627, -0.03300013020634651, 0.007833226583898067, -0.03984469920396805, 0.0315018855035305, 0.009249984286725521, -0.0127114...
1,560
1,560
['Ehsaneddin Asgari', 'Hinrich Schütze']
1704.08914v2
We present SuperPivot, an analysis method for low-resource languages that occur in a superparallel corpus, i.e., in a corpus that contains an order of magnitude more languages than parallel corpora currently in use. We show that SuperPivot performs well for the crosslingual analysis of the linguistic phenomenon of tens...
Past, Present, Future: A Computational Investigation of the Typology of Tense in 1000 Languages
2,017
http://arxiv.org/pdf/1704.08914v2
Title Past Present Future Computational Investigation Typology Tense 1000 Languages Summary present SuperPivot analysis method lowresource language occur superparallel corpus ie corpus contains order magnitude language parallel corpus currently use show SuperPivot performs well crosslingual analysis linguistic phenomen...
[0.021771324798464775, -0.012175197713077068, -0.04369158297777176, 0.04584728553891182, -0.03316947817802429, 0.023444680497050285, 0.03758956491947174, 0.011688983999192715, -0.010391401126980782, -0.05489262565970421, 0.015437053516507149, -0.04574887454509735, 0.064653180539608, -0.013435198925435543, 0.02388496510...
1,561
1,561
['Ankur Bapna', 'Gokhan Tur', 'Dilek Hakkani-Tur', 'Larry Heck']
1705.03455v3
Spoken Language Understanding (SLU) is a key component of goal oriented dialogue systems that would parse user utterances into semantic frame representations. Traditionally SLU does not utilize the dialogue history beyond the previous system turn and contextual ambiguities are resolved by the downstream components. In ...
Sequential Dialogue Context Modeling for Spoken Language Understanding
2,017
http://arxiv.org/pdf/1705.03455v3
Title Sequential Dialogue Context Modeling Spoken Language Understanding Summary Spoken Language Understanding SLU key component goal oriented dialogue system would parse user utterance semantic frame representation Traditionally SLU utilize dialogue history beyond previous system turn contextual ambiguity resolved dow...
[0.04887152090668678, 0.03807653486728668, 0.009634042158722878, 0.07587263733148575, -0.005208486225455999, -0.00043677393114194274, 0.025058498606085777, -0.034408826380968094, -0.03842048719525337, -0.043647799640893936, 0.00047181607806123793, -0.05209955945611, 0.027239514514803886, 0.04976595193147659, -0.0491305...
1,562
1,562
['Qizhe Xie', 'Zihang Dai', 'Yulun Du', 'Eduard Hovy', 'Graham Neubig']
1705.11122v3
Learning meaningful representations that maintain the content necessary for a particular task while filtering away detrimental variations is a problem of great interest in machine learning. In this paper, we tackle the problem of learning representations invariant to a specific factor or trait of data. The representati...
Controllable Invariance through Adversarial Feature Learning
2,017
http://arxiv.org/pdf/1705.11122v3
Title Controllable Invariance Adversarial Feature Learning Summary Learning meaningful representation maintain content necessary particular task filtering away detrimental variation problem great interest machine learning paper tackle problem learning representation invariant specific factor trait data representation l...
[0.01575343869626522, 0.03383159264922142, -0.017317963764071465, 0.0367632731795311, -0.015838351100683212, -0.0027708595152944326, 0.02420765720307827, -0.005822396371513605, -0.005358574911952019, -0.03464142978191376, -0.015820695087313652, 0.028130479156970978, -0.0303848497569561, 0.042719557881355286, 0.03510127...
1,563
1,563
['Nikola Mrkšić', 'Ivan Vulić', 'Diarmuid Ó Séaghdha', 'Ira Leviant', 'Roi Reichart', 'Milica Gašić', 'Anna Korhonen', 'Steve Young']
1706.00374v1
We present Attract-Repel, an algorithm for improving the semantic quality of word vectors by injecting constraints extracted from lexical resources. Attract-Repel facilitates the use of constraints from mono- and cross-lingual resources, yielding semantically specialised cross-lingual vector spaces. Our evaluation show...
Semantic Specialisation of Distributional Word Vector Spaces using Monolingual and Cross-Lingual Constraints
2,017
http://arxiv.org/pdf/1706.00374v1
Title Semantic Specialisation Distributional Word Vector Spaces using Monolingual CrossLingual Constraints Summary present AttractRepel algorithm improving semantic quality word vector injecting constraint extracted lexical resource AttractRepel facilitates use constraint mono crosslingual resource yielding semanticall...
[0.05038539320230484, 0.015305785462260246, -0.01764019951224327, 0.06330706924200058, -0.03240390866994858, -0.010728398337960243, -0.025464342907071114, -0.005344782955944538, -0.03214089199900627, -0.04832526668906212, -0.04580532759428024, -0.017099130898714066, 0.0016986323753371835, -0.0066307466477155685, 0.0230...
1,564
1,564
['Pei-Hao Su', 'Pawel Budzianowski', 'Stefan Ultes', 'Milica Gasic', 'Steve Young']
1707.00130v2
Deep reinforcement learning (RL) methods have significant potential for dialogue policy optimisation. However, they suffer from a poor performance in the early stages of learning. This is especially problematic for on-line learning with real users. Two approaches are introduced to tackle this problem. Firstly, to speed...
Sample-efficient Actor-Critic Reinforcement Learning with Supervised Data for Dialogue Management
2,017
http://arxiv.org/pdf/1707.00130v2
Title Sampleefficient ActorCritic Reinforcement Learning Supervised Data Dialogue Management Summary Deep reinforcement learning RL method significant potential dialogue policy optimisation However suffer poor performance early stage learning especially problematic online learning real user Two approach introduced tack...
[0.07835343480110168, 0.07937835156917572, -0.009527315385639668, 0.014157083816826344, 0.00016248185420408845, 0.0005534474039450288, 0.011633096262812614, -0.0007159101660363376, -0.022400712594389915, -0.0386226512491703, -0.029218120500445366, -0.01053797360509634, -0.038305170834064484, 0.07966911792755127, -0.044...
1,565
1,565
['Edilson A. Corrêa Jr', 'Vanessa Q. Marinho', 'Leandro B. dos Santos', 'Thales F. C. Bertaglia', 'Marcos V. Treviso', 'Henrico B. Brum']
1707.02657v1
The enormous amount of texts published daily by Internet users has fostered the development of methods to analyze this content in several natural language processing areas, such as sentiment analysis. The main goal of this task is to classify the polarity of a message. Even though many approaches have been proposed for...
PELESent: Cross-domain polarity classification using distant supervision
2,017
http://arxiv.org/pdf/1707.02657v1
Title PELESent Crossdomain polarity classification using distant supervision Summary enormous amount text published daily Internet user fostered development method analyze content several natural language processing area sentiment analysis main goal task classify polarity message Even though many approach proposed sent...
[0.0584435872733593, 0.03217697888612747, -0.021723248064517975, 0.01272839680314064, -0.02245350554585457, 0.04096490517258644, 0.0273430198431015, 0.018364833667874336, 0.024320829659700394, -0.10186731070280075, -0.027870576828718185, 0.0010069457348436117, -0.03267904371023178, 0.02207578346133232, -0.0389609858393...
1,566
1,566
['Hao Tan', 'Mohit Bansal']
1707.03804v2
Models that can execute natural language instructions for situated robotic tasks such as assembly and navigation have several useful applications in homes, offices, and remote scenarios. We study the semantics of spatially-referred configuration and arrangement instructions, based on the challenging Bisk-2016 blank-lab...
Source-Target Inference Models for Spatial Instruction Understanding
2,017
http://arxiv.org/pdf/1707.03804v2
Title SourceTarget Inference Models Spatial Instruction Understanding Summary Models execute natural language instruction situated robotic task assembly navigation several useful application home office remote scenario study semantics spatiallyreferred configuration arrangement instruction based challenging Bisk2016 bl...
[0.05101113021373749, 0.01423615776002407, 0.004879285115748644, 0.0021390924230217934, -0.012735840864479542, -0.008939577266573906, 0.07184893637895584, -0.040301363915205, -0.026051389053463936, -0.07678220421075821, -0.009330382570624352, 0.020752068608999252, 0.017685722559690475, -0.0032553013879805803, 0.0306806...
1,567
1,567
['Khanh Nguyen', 'Hal Daumé III', 'Jordan Boyd-Graber']
1707.07402v4
Machine translation is a natural candidate problem for reinforcement learning from human feedback: users provide quick, dirty ratings on candidate translations to guide a system to improve. Yet, current neural machine translation training focuses on expensive human-generated reference translations. We describe a reinfo...
Reinforcement Learning for Bandit Neural Machine Translation with Simulated Human Feedback
2,017
http://arxiv.org/pdf/1707.07402v4
Title Reinforcement Learning Bandit Neural Machine Translation Simulated Human Feedback Summary Machine translation natural candidate problem reinforcement learning human feedback user provide quick dirty rating candidate translation guide system improve Yet current neural machine translation training focus expensive h...
[0.0757022425532341, 0.020084230229258537, -0.010482442565262318, -0.005405619274824858, -0.0008842404931783676, 0.017441706731915474, -0.006689753849059343, 0.022122010588645935, -0.03102836385369301, -0.02633756957948208, -0.013218934647738934, -0.025616100057959557, 0.021547852084040642, 0.030500030145049095, 0.0404...
1,568
1,568
['Mostafa Dehghani', 'Hosein Azarbonyad', 'Jaap Kamps', 'Maarten de Rijke']
1707.07605v1
Deep neural networks have become a primary tool for solving problems in many fields. They are also used for addressing information retrieval problems and show strong performance in several tasks. Training these models requires large, representative datasets and for most IR tasks, such data contains sensitive informatio...
Share your Model instead of your Data: Privacy Preserving Mimic Learning for Ranking
2,017
http://arxiv.org/pdf/1707.07605v1
Title Share Model instead Data Privacy Preserving Mimic Learning Ranking Summary Deep neural network become primary tool solving problem many field also used addressing information retrieval problem show strong performance several task Training model requires large representative datasets IR task data contains sensitiv...
[0.025296255946159363, 0.09064938127994537, -0.03974328190088272, 0.01920529082417488, -0.03810644894838333, -0.018941940739750862, 0.08087491244077682, -0.013321487233042717, 0.018488595262169838, -0.007602659519761801, -0.07811205834150314, -0.01131855882704258, -0.057737842202186584, 0.07120604068040848, -0.01620232...
1,569
1,569
['Lei Sha', 'Lili Mou', 'Tianyu Liu', 'Pascal Poupart', 'Sujian Li', 'Baobao Chang', 'Zhifang Sui']
1709.00155v1
Generating texts from structured data (e.g., a table) is important for various natural language processing tasks such as question answering and dialog systems. In recent studies, researchers use neural language models and encoder-decoder frameworks for table-to-text generation. However, these neural network-based appro...
Order-Planning Neural Text Generation From Structured Data
2,017
http://arxiv.org/pdf/1709.00155v1
Title OrderPlanning Neural Text Generation Structured Data Summary Generating text structured data eg table important various natural language processing task question answering dialog system recent study researcher use neural language model encoderdecoder framework tabletotext generation However neural networkbased ap...
[0.041281577199697495, 0.09511860460042953, -0.009292027913033962, 0.027168555185198784, -0.005965259857475758, 0.011682846583425999, 0.013681650161743164, -0.0029423097148537636, -0.012972340919077396, -0.028641484677791595, 0.04243650287389755, -0.05520966649055481, 0.03221803158521652, 0.07633661478757858, 0.0481963...
1,570
1,570
['Tianyu Li', 'Guillaume Rabusseau', 'Doina Precup']
1709.04380v2
Weighted finite automata (WFA) can expressively model functions defined over strings but are inherently linear models. Given the recent successes of nonlinear models in machine learning, it is natural to wonder whether ex-tending WFA to the nonlinear setting would be beneficial. In this paper, we propose a novel model ...
Neural Network Based Nonlinear Weighted Finite Automata
2,017
http://arxiv.org/pdf/1709.04380v2
Title Neural Network Based Nonlinear Weighted Finite Automata Summary Weighted finite automaton WFA expressively model function defined string inherently linear model Given recent success nonlinear model machine learning natural wonder whether extending WFA nonlinear setting would beneficial paper propose novel model n...
[0.02375919371843338, 0.09754817932844162, -0.01853114180266857, 0.019844945520162582, -0.009679291397333145, -0.01414862833917141, -0.036672841757535934, 0.014087089337408543, -0.021719731390476227, -0.03648659586906433, -0.017837516963481903, -0.01310208160430193, 0.005114678759127855, 0.049783166497945786, 0.0348960...
1,571
1,571
['Yanchao Yu', 'Arash Eshghi', 'Oliver Lemon']
1709.10423v1
We present an optimised multi-modal dialogue agent for interactive learning of visually grounded word meanings from a human tutor, trained on real human-human tutoring data. Within a life-long interactive learning period, the agent, trained using Reinforcement Learning (RL), must be able to handle natural conversations...
Learning how to learn: an adaptive dialogue agent for incrementally learning visually grounded word meanings
2,017
http://arxiv.org/pdf/1709.10423v1
Title Learning learn adaptive dialogue agent incrementally learning visually grounded word meaning Summary present optimised multimodal dialogue agent interactive learning visually grounded word meaning human tutor trained real humanhuman tutoring data Within lifelong interactive learning period agent trained using Rei...
[0.06880705058574677, 0.02616000361740589, -0.009797616861760616, 0.0394420251250267, -0.010853959247469902, -0.0029809519182890654, 0.012037869542837143, 0.0014049011515453458, -0.015049920417368412, -0.06258079409599304, -0.016840092837810516, 0.016925357282161713, 1.3654121175932232e-05, 0.07815419137477875, -0.0032...
1,572
1,572
['Brenden M. Lake', 'Marco Baroni']
1711.00350v2
Humans can understand and produce new utterances effortlessly, thanks to their compositional skills. Once a person learns the meaning of a new verb "dax," he or she can immediately understand the meaning of "dax twice" or "sing and dax." In this paper, we introduce the SCAN domain, consisting of a set of simple composi...
Generalization without systematicity: On the compositional skills of sequence-to-sequence recurrent networks
2,017
http://arxiv.org/pdf/1711.00350v2
Title Generalization without systematicity compositional skill sequencetosequence recurrent network Summary Humans understand produce new utterance effortlessly thanks compositional skill person learns meaning new verb dax immediately understand meaning dax twice sing dax paper introduce SCAN domain consisting set simp...
[0.025033805519342422, 0.08294496685266495, -0.019171910360455513, -0.0077891587279737, -0.01290016807615757, 0.01754045858979225, 0.017018716782331467, 0.0003580524353310466, -0.10141097009181976, -0.011952889151871204, -0.01293276622891426, -0.04158027470111847, 0.046497754752635956, 0.0420285165309906, 0.05632517859...
1,573
1,573
['Omer Goldman', 'Veronica Latcinnik', 'Udi Naveh', 'Amir Globerson', 'Jonathan Berant']
1711.05240v1
Semantic parsers translate language utterances to programs, but are often trained from utterance-denotation pairs only. Consequently, parsers must overcome the problem of spuriousness at training time, where an incorrect program found at search time accidentally leads to a correct denotation. We propose that in small w...
Weakly-supervised Semantic Parsing with Abstract Examples
2,017
http://arxiv.org/pdf/1711.05240v1
Title Weaklysupervised Semantic Parsing Abstract Examples Summary Semantic parser translate language utterance program often trained utterancedenotation pair Consequently parser must overcome problem spuriousness training time incorrect program found search time accidentally lead correct denotation propose small wellty...
[0.05658732354640961, 0.019172145053744316, -0.003598341252654791, 0.05037634074687958, -0.029716182500123978, 0.0193550456315279, 0.0002946644090116024, -0.0391303151845932, -0.0035328108351677656, -0.0692300871014595, -0.010336233302950859, 0.03692746162414551, -0.02028759755194187, 0.044351350516080856, 0.0047738216...
1,574
1,574
['Zachary Lipton', 'Xiujun Li', 'Jianfeng Gao', 'Lihong Li', 'Faisal Ahmed', 'Li Deng']
1711.05715v2
We present a new algorithm that significantly improves the efficiency of exploration for deep Q-learning agents in dialogue systems. Our agents explore via Thompson sampling, drawing Monte Carlo samples from a Bayes-by-Backprop neural network. Our algorithm learns much faster than common exploration strategies such as ...
BBQ-Networks: Efficient Exploration in Deep Reinforcement Learning for Task-Oriented Dialogue Systems
2,017
http://arxiv.org/pdf/1711.05715v2
Title BBQNetworks Efficient Exploration Deep Reinforcement Learning TaskOriented Dialogue Systems Summary present new algorithm significantly improves efficiency exploration deep Qlearning agent dialogue system agent explore via Thompson sampling drawing Monte Carlo sample BayesbyBackprop neural network algorithm learn...
[0.01756395399570465, -0.002453365596011281, -0.006325583439320326, 0.007751567754894495, 0.0036683231592178345, 0.004726892337203026, -0.002970527857542038, -0.003614576067775488, -0.025236308574676514, -0.012071250006556511, -0.06712344288825989, 0.004469552543014288, -0.030904719606041908, 0.08419699221849442, -0.01...
1,575
1,575
['Amir Zadeh', 'Paul Pu Liang', 'Soujanya Poria', 'Prateek Vij', 'Erik Cambria', 'Louis-Philippe Morency']
1802.00923v1
Human face-to-face communication is a complex multimodal signal. We use words (language modality), gestures (vision modality) and changes in tone (acoustic modality) to convey our intentions. Humans easily process and understand face-to-face communication, however, comprehending this form of communication remains a sig...
Multi-attention Recurrent Network for Human Communication Comprehension
2,018
http://arxiv.org/pdf/1802.00923v1
Title Multiattention Recurrent Network Human Communication Comprehension Summary Human facetoface communication complex multimodal signal use word language modality gesture vision modality change tone acoustic modality convey intention Humans easily process understand facetoface communication however comprehending form...
[0.028437985107302666, 0.027803272008895874, -0.026736563071608543, 0.03625517338514328, -0.017403123900294304, -0.002322165062651038, 0.011146239005029202, -0.008715088479220867, -0.007865844294428825, -0.026071244850754738, -0.038264285773038864, -0.06105484440922737, 0.055823445320129395, 0.07047637552022934, 0.0222...
1,576
1,576
['Haonan Yu', 'Haichao Zhang', 'Wei Xu']
1802.01433v2
We build a virtual agent for learning language in a 2D maze-like world. The agent sees images of the surrounding environment, listens to a virtual teacher, and takes actions to receive rewards. It interactively learns the teacher's language from scratch based on two language use cases: sentence-directed navigation and ...
Interactive Grounded Language Acquisition and Generalization in a 2D World
2,018
http://arxiv.org/pdf/1802.01433v2
Title Interactive Grounded Language Acquisition Generalization 2D World Summary build virtual agent learning language 2D mazelike world agent see image surrounding environment listens virtual teacher take action receive reward interactively learns teacher language scratch based two language use case sentencedirected na...
[0.04892810806632042, 0.0017932787304744124, -0.0289580337703228, 0.056983765214681625, -0.016123315319418907, 0.012060612440109253, -0.022748388350009918, -0.003439185908064246, -0.0091802217066288, -0.04148978739976883, -0.03195532038807869, 0.046840012073516846, -0.009573470801115036, 0.05904536694288254, 0.01355563...
1,577
1,577
['Junyang Lin', 'Shuming Ma', 'Qi Su', 'Xu Sun']
1802.01812v1
Attention-based sequence-to-sequence model has proved successful in Neural Machine Translation (NMT). However, the attention without consideration of decoding history, which includes the past information in the decoder and the attention mechanism, often causes much repetition. To address this problem, we propose the de...
Decoding-History-Based Adaptive Control of Attention for Neural Machine Translation
2,018
http://arxiv.org/pdf/1802.01812v1
Title DecodingHistoryBased Adaptive Control Attention Neural Machine Translation Summary Attentionbased sequencetosequence model proved successful Neural Machine Translation NMT However attention without consideration decoding history includes past information decoder attention mechanism often cause much repetition add...
[0.053714074194431305, 0.04717693850398064, -0.018873877823352814, 0.034741126000881195, -0.015201386995613575, 0.02089119143784046, -0.006957128178328276, -0.007472123950719833, -0.006572692189365625, 0.004302085377275944, 0.0013095242902636528, -0.06890688836574554, 0.03302856162190437, 0.009929809719324112, 0.002472...
1,578
1,578
['Yelong Shen', 'Jianshu Chen', 'Po-Sen Huang', 'Yuqing Guo', 'Jianfeng Gao']
1802.04394v1
Learning to walk over a graph towards a target node for a given input query and a source node is an important problem in applications such as knowledge graph reasoning. It can be formulated as a reinforcement learning (RL) problem that has a known state transition model, but with partial observability and sparse reward...
ReinforceWalk: Learning to Walk in Graph with Monte Carlo Tree Search
2,018
http://arxiv.org/pdf/1802.04394v1
Title ReinforceWalk Learning Walk Graph Monte Carlo Tree Search Summary Learning walk graph towards target node given input query source node important problem application knowledge graph reasoning formulated reinforcement learning RL problem known state transition model partial observability sparse reward overcome cha...
[0.015647483989596367, 0.014903712086379528, 0.002254936145618558, -0.06005411222577095, -0.031569112092256546, -0.06113411486148834, -0.0629810094833374, -0.0036915745586156845, -0.011914744041860104, 0.017698809504508972, 0.037249285727739334, 0.04163502901792526, -0.01455253642052412, 0.061114974319934845, 0.0240776...
1,579
1,579
['Shaojie Bai', 'J. Zico Kolter', 'Vladlen Koltun']
1803.01271v1
For most deep learning practitioners, sequence modeling is synonymous with recurrent networks. Yet recent results indicate that convolutional architectures can outperform recurrent networks on tasks such as audio synthesis and machine translation. Given a new sequence modeling task or dataset, which architecture should...
An Empirical Evaluation of Generic Convolutional and Recurrent Networks for Sequence Modeling
2,018
http://arxiv.org/pdf/1803.01271v1
Title Empirical Evaluation Generic Convolutional Recurrent Networks Sequence Modeling Summary deep learning practitioner sequence modeling synonymous recurrent network Yet recent result indicate convolutional architecture outperform recurrent network task audio synthesis machine translation Given new sequence modeling ...
[0.035510044544935226, 0.010446072556078434, -0.011199641041457653, 0.03390893712639809, 0.0005593755631707609, -0.02324017509818077, 0.051391419023275375, -0.020214248448610306, -0.0680561512708664, -0.0003581522614695132, -0.018302420154213905, -0.03833095729351044, 0.03442465886473656, 0.041028816252946854, 0.018001...
1,580
1,580
['Ido Dagan', 'Lillian Lee', 'Fernando C. N. Pereira']
cs/9809110v1
In many applications of natural language processing (NLP) it is necessary to determine the likelihood of a given word combination. For example, a speech recognizer may need to determine which of the two word combinations ``eat a peach'' and ``eat a beach'' is more likely. Statistical NLP methods determine the likelihoo...
Similarity-Based Models of Word Cooccurrence Probabilities
1,998
http://arxiv.org/pdf/cs/9809110v1
Title SimilarityBased Models Word Cooccurrence Probabilities Summary many application natural language processing NLP necessary determine likelihood given word combination example speech recognizer may need determine two word combination eat peach eat beach likely Statistical NLP method determine likelihood word combin...
[0.04290740191936493, 0.01982738822698593, 0.007747357711195946, 0.0541904978454113, -0.058120131492614746, -0.002541005378589034, -0.011427572928369045, 0.005639173556119204, -0.018292155116796494, -0.056601639837026596, -0.04131583496928215, -0.011025008745491505, 0.07181946933269501, 0.015346693806350231, 0.00456168...
1,581
1,581
['Oren Glickman', 'Ido Dagan']
cs/0312058v1
This paper studies the potential of identifying lexical paraphrases within a single corpus, focusing on the extraction of verb paraphrases. Most previous approaches detect individual paraphrase instances within a pair (or set) of comparable corpora, each of them containing roughly the same information, and rely on the ...
Acquiring Lexical Paraphrases from a Single Corpus
2,003
http://arxiv.org/pdf/cs/0312058v1
Title Acquiring Lexical Paraphrases Single Corpus Summary paper study potential identifying lexical paraphrase within single corpus focusing extraction verb paraphrase previous approach detect individual paraphrase instance within pair set comparable corpus containing roughly information rely substantial level correspo...
[0.08003520220518112, 0.013254516758024693, -0.00900109950453043, 0.05898572504520416, -0.06599742919206619, -0.0005100962589494884, 0.014274165965616703, 0.04246765747666359, -0.04891467094421387, -0.0662238821387291, 0.013013503514230251, 0.043244995176792145, 0.01685011200606823, -0.0563029982149601, -0.005571264773...
1,582
1,582
['Peter D. Turney']
cs/0607120v1
We present an unsupervised learning algorithm that mines large text corpora for patterns that express implicit semantic relations. For a given input word pair X:Y with some unspecified semantic relations, the corresponding output list of patterns <P1,...,Pm> is ranked according to how well each pattern Pi expresses the...
Expressing Implicit Semantic Relations without Supervision
2,006
http://arxiv.org/pdf/cs/0607120v1
Title Expressing Implicit Semantic Relations without Supervision Summary present unsupervised learning algorithm mine large text corpus pattern express implicit semantic relation given input word pair XY unspecified semantic relation corresponding output list pattern P1Pm ranked according well pattern Pi express relati...
[0.05976976454257965, 0.060216449201107025, -0.011576998047530651, 0.06998075544834137, -0.033780813217163086, -0.00445672357454896, 0.0068848454393446445, 0.03983183577656746, 0.013157999143004417, -0.06719060242176056, 0.0020354739390313625, 0.06836274266242981, 0.0012908705975860357, 0.019134750589728355, -0.0004579...
1,583
1,583
['Sergio Guadarrama', 'David P. Pancho']
1005.5253v1
The contribution of this paper is to provide a semantic model (using soft constraints) of the words used by web-users to describe objects in a language game; a game in which one user describes a selected object of those composing the scene, and another user has to guess which object has been described. The given descri...
Using Soft Constraints To Learn Semantic Models Of Descriptions Of Shapes
2,010
http://arxiv.org/pdf/1005.5253v1
Title Using Soft Constraints Learn Semantic Models Descriptions Shapes Summary contribution paper provide semantic model using soft constraint word used webusers describe object language game game one user describes selected object composing scene another user guess object described given description need non ambiguous...
[0.06746505200862885, 0.03244669735431671, -0.013010627590119839, 0.061777301132678986, -0.01654740236699581, -0.013407536782324314, 0.0005610708612948656, 0.007002146914601326, -0.0636228397488594, -0.07747069001197815, -0.0044388724491000175, 0.01123740617185831, -0.0005609633517451584, 0.0963551476597786, -0.0184608...
1,584
1,584
['Stephen Jose Hanson', 'Malcolm Bauer']
1304.3432v1
This paper describes a machine induction program (WITT) that attempts to model human categorization. Properties of categories to which human subjects are sensitive includes best or prototypical members, relative contrasts between putative categories, and polymorphy (neither necessary or sufficient features). This appro...
Machine Learning, Clustering, and Polymorphy
2,013
http://arxiv.org/pdf/1304.3432v1
Title Machine Learning Clustering Polymorphy Summary paper describes machine induction program WITT attempt model human categorization Properties category human subject sensitive includes best prototypical member relative contrast putative category polymorphy neither necessary sufficient feature approach represents alt...
[0.012280922383069992, -0.022099941968917847, -0.03277726471424103, 0.0008489289903081954, 0.01642909273505211, 0.020806316286325455, 0.02917058765888214, 0.011869069188833237, 0.023669075220823288, -0.03962359204888344, 0.01286787074059248, 0.023881932720541954, 0.0016144413966685534, 0.0021279482170939445, 0.03238074...
1,585
1,585
['Catherine Recanati', 'Nicoleta Rogovschi', 'Younès Bennani']
0710.2446v1
Data mining allows the exploration of sequences of phenomena, whereas one usually tends to focus on isolated phenomena or on the relation between two phenomena. It offers invaluable tools for theoretical analyses and exploration of the structure of sentences, texts, dialogues, and speech. We report here the results of ...
The structure of verbal sequences analyzed with unsupervised learning techniques
2,007
http://arxiv.org/pdf/0710.2446v1
Title structure verbal sequence analyzed unsupervised learning technique Summary Data mining allows exploration sequence phenomenon whereas one usually tends focus isolated phenomenon relation two phenomenon offer invaluable tool theoretical analysis exploration structure sentence text dialogue speech report result att...
[0.0057299197651445866, 0.03894606605172157, -0.028538495302200317, 0.06058136373758316, -0.04934658110141754, 0.016575250774621964, -0.006524731405079365, 0.03017023392021656, 0.00044814302236773074, -0.06881996244192123, 0.048590660095214844, 0.00522107258439064, 0.038056593388319016, 0.04577933996915817, -0.00848710...
1,586
1,586
['Peter D. Turney']
1107.4573v1
It has been argued that analogy is the core of cognition. In AI research, algorithms for analogy are often limited by the need for hand-coded high-level representations as input. An alternative approach is to use high-level perception, in which high-level representations are automatically generated from raw data. Analo...
Analogy perception applied to seven tests of word comprehension
2,011
http://arxiv.org/pdf/1107.4573v1
Title Analogy perception applied seven test word comprehension Summary argued analogy core cognition AI research algorithm analogy often limited need handcoded highlevel representation input alternative approach use highlevel perception highlevel representation automatically generated raw data Analogy perception proces...
[0.046695832163095474, 0.018182754516601562, -0.01578073762357235, 0.05215299502015114, -0.05495433881878853, 0.030628638342022896, -0.0034256940707564354, 0.003073861123993993, -0.011637278832495213, -0.04211779683828354, -0.013853340409696102, 0.017886795103549957, 0.048689328134059906, -0.028279908001422882, -0.0128...
1,587
1,587
['Peter D. Turney']
0812.4446v1
Many AI researchers and cognitive scientists have argued that analogy is the core of cognition. The most influential work on computational modeling of analogy-making is Structure Mapping Theory (SMT) and its implementation in the Structure Mapping Engine (SME). A limitation of SME is the requirement for complex hand-co...
The Latent Relation Mapping Engine: Algorithm and Experiments
2,008
http://arxiv.org/pdf/0812.4446v1
Title Latent Relation Mapping Engine Algorithm Experiments Summary Many AI researcher cognitive scientist argued analogy core cognition influential work computational modeling analogymaking Structure Mapping Theory SMT implementation Structure Mapping Engine SME limitation SME requirement complex handcoded representati...
[0.05943764001131058, 0.03560594096779823, -0.013670702464878559, 0.03014051727950573, -0.03500904515385628, 0.03775309771299362, -0.016835469752550125, -0.0017366887768730521, 0.008360047824680805, -0.06223591789603233, -0.004695680923759937, 0.0266692116856575, 0.015125851146876812, 0.014485207386314869, -0.001705809...
1,588
1,588
['Guntis Barzdins']
1406.2538v1
The paper presents a FrameNet-based information extraction and knowledge representation framework, called FrameNet-CNL. The framework is used on natural language documents and represents the extracted knowledge in a tailor-made Frame-ontology from which unambiguous FrameNet-CNL paraphrase text can be generated automati...
FrameNet CNL: a Knowledge Representation and Information Extraction Language
2,014
http://arxiv.org/pdf/1406.2538v1
Title FrameNet CNL Knowledge Representation Information Extraction Language Summary paper present FrameNetbased information extraction knowledge representation framework called FrameNetCNL framework used natural language document represents extracted knowledge tailormade Frameontology unambiguous FrameNetCNL paraphrase...
[0.024309681728482246, 0.02717193402349949, 0.0019784600008279085, 0.06157669425010681, -0.062458187341690063, 0.011010616086423397, 0.0089713204652071, 0.026845242828130722, -0.004956433083862066, -0.07505645602941513, 0.021716641262173653, 0.053256675601005554, -0.0068820477463305, 0.03314409404993057, -0.01057983189...
1,589
1,589
['Marc Dymetman', 'Guillaume Bouchard', 'Simon Carter']
1207.0742v1
Most current sampling algorithms for high-dimensional distributions are based on MCMC techniques and are approximate in the sense that they are valid only asymptotically. Rejection sampling, on the other hand, produces valid samples, but is unrealistically slow in high-dimension spaces. The OS* algorithm that we propos...
The OS* Algorithm: a Joint Approach to Exact Optimization and Sampling
2,012
http://arxiv.org/pdf/1207.0742v1
Title OS Algorithm Joint Approach Exact Optimization Sampling Summary current sampling algorithm highdimensional distribution based MCMC technique approximate sense valid asymptotically Rejection sampling hand produce valid sample unrealistically slow highdimension space OS algorithm propose unified approach exact opti...
[-0.012311781756579876, 0.03389445319771767, -0.008573406375944614, -0.00623664353042841, -0.035678088665008545, -0.015572474338114262, -0.003842942649498582, 0.006077030673623085, -0.02133714035153389, -0.012374830432236195, 0.018265729770064354, 0.00320060970261693, 0.0027076348196715117, 0.003933027386665344, 0.0069...
1,590
1,590
['Dana Movshovitz-Attias', 'William W. Cohen']
1505.00277v1
We present an approach for the detection of coordinate-term relationships between entities from the software domain, that refer to Java classes. Usually, relations are found by examining corpus statistics associated with text entities. In some technical domains, however, we have access to additional information about t...
Grounded Discovery of Coordinate Term Relationships between Software Entities
2,015
http://arxiv.org/pdf/1505.00277v1
Title Grounded Discovery Coordinate Term Relationships Software Entities Summary present approach detection coordinateterm relationship entity software domain refer Java class Usually relation found examining corpus statistic associated text entity technical domain however access additional information realworld object...
[0.06125997379422188, 0.024601202458143234, -0.019413761794567108, 0.04472612962126732, -0.014832010492682457, 0.002859273925423622, 0.028782740235328674, 0.03412811830639839, -0.002692160662263632, -0.054024990648031235, -0.0244210883975029, 0.018205493688583374, 0.06187310069799423, 0.009449012577533722, -0.030151203...
1,591
1,591
['Matthew R. Gormley', 'Mo Yu', 'Mark Dredze']
1505.02419v3
Compositional embedding models build a representation (or embedding) for a linguistic structure based on its component word embeddings. We propose a Feature-rich Compositional Embedding Model (FCM) for relation extraction that is expressive, generalizes to new domains, and is easy-to-implement. The key idea is to combi...
Improved Relation Extraction with Feature-Rich Compositional Embedding Models
2,015
http://arxiv.org/pdf/1505.02419v3
Title Improved Relation Extraction FeatureRich Compositional Embedding Models Summary Compositional embedding model build representation embedding linguistic structure based component word embeddings propose Featurerich Compositional Embedding Model FCM relation extraction expressive generalizes new domain easytoimplem...
[0.048786263912916183, 0.031159156933426857, 0.0038183056749403477, 0.08961445838212967, -0.023788921535015106, 0.01219424232840538, 0.00022908060054760426, -0.009485345333814621, 0.0005338096525520086, -0.04486526921391487, -0.02120903879404068, -0.012306987307965755, 0.018685150891542435, 0.01543318573385477, -0.0162...
1,592
1,592
['David Golub', 'Xiaodong He']
1604.00727v4
We show that a character-level encoder-decoder framework can be successfully applied to question answering with a structured knowledge base. We use our model for single-relation question answering and demonstrate the effectiveness of our approach on the SimpleQuestions dataset (Bordes et al., 2015), where we improve st...
Character-Level Question Answering with Attention
2,016
http://arxiv.org/pdf/1604.00727v4
Title CharacterLevel Question Answering Attention Summary show characterlevel encoderdecoder framework successfully applied question answering structured knowledge base use model singlerelation question answering demonstrate effectiveness approach SimpleQuestions dataset Bordes et al 2015 improve stateoftheart accuracy...
[0.06037707254290581, 0.05398186296224594, 0.013545789755880833, 0.03059883788228035, 0.012521451339125633, 0.04000189155340195, -4.2515599488979205e-05, -7.880829798523337e-05, -0.008750548586249352, -0.023072823882102966, 0.022018788382411003, -0.024119677022099495, -0.018705036491155624, 0.05458735674619675, 0.03272...
1,593
1,593
['Quoc V. Le', 'Tomas Mikolov']
1405.4053v2
Many machine learning algorithms require the input to be represented as a fixed-length feature vector. When it comes to texts, one of the most common fixed-length features is bag-of-words. Despite their popularity, bag-of-words features have two major weaknesses: they lose the ordering of the words and they also ignore...
Distributed Representations of Sentences and Documents
2,014
http://arxiv.org/pdf/1405.4053v2
Title Distributed Representations Sentences Documents Summary Many machine learning algorithm require input represented fixedlength feature vector come text one common fixedlength feature bagofwords Despite popularity bagofwords feature two major weakness lose ordering word also ignore semantics word example powerful s...
[0.03824855387210846, 0.044329285621643066, 0.0028755615931004286, 0.04347148537635803, -0.05934135988354683, 0.02382882684469223, 0.03260527551174164, -0.0021368355955928564, 0.0072017936035990715, -0.10406472533941269, -0.007800460327416658, 0.018961479887366295, 0.008516795001924038, 0.027384771034121513, 0.00188498...
1,594
1,594
['Peter D. Turney']
1405.7908v1
Semantic composition is the task of understanding the meaning of text by composing the meanings of the individual words in the text. Semantic decomposition is the task of understanding the meaning of an individual word by decomposing it into various aspects (factors, constituents, components) that are latent in the mea...
Semantic Composition and Decomposition: From Recognition to Generation
2,014
http://arxiv.org/pdf/1405.7908v1
Title Semantic Composition Decomposition Recognition Generation Summary Semantic composition task understanding meaning text composing meaning individual word text Semantic decomposition task understanding meaning individual word decomposing various aspect factor constituent component latent meaning word take distribut...
[0.05148473009467125, 0.05103696882724762, -5.074320870335214e-05, 0.06901339441537857, -0.037829626351594925, 0.02127276174724102, 0.014219788834452629, 0.03459308296442032, -0.06276735663414001, -0.06117156893014908, 0.0036030288320034742, 0.039595991373062134, 0.03444637358188629, -0.01428279560059309, -0.0253729447...
1,595
1,595
['Kai Sheng Tai', 'Richard Socher', 'Christopher D. Manning']
1503.00075v3
Because of their superior ability to preserve sequence information over time, Long Short-Term Memory (LSTM) networks, a type of recurrent neural network with a more complex computational unit, have obtained strong results on a variety of sequence modeling tasks. The only underlying LSTM structure that has been explored...
Improved Semantic Representations From Tree-Structured Long Short-Term Memory Networks
2,015
http://arxiv.org/pdf/1503.00075v3
Title Improved Semantic Representations TreeStructured Long ShortTerm Memory Networks Summary superior ability preserve sequence information time Long ShortTerm Memory LSTM network type recurrent neural network complex computational unit obtained strong result variety sequence modeling task underlying LSTM structure ex...
[0.018036363646388054, 0.014399793930351734, -0.012372614815831184, 0.05039801076054573, -0.042778752744197845, 0.005895053036510944, -0.0517454519867897, 0.017162395641207695, -0.028207343071699142, -0.06820531189441681, -0.01547558419406414, -0.0034219003282487392, 0.02757977321743965, 0.016290418803691864, -0.013350...
1,596
1,596
['Biao Liu', 'Minlie Huang']
1503.00841v1
Prior knowledge has been shown very useful to address many natural language processing tasks. Many approaches have been proposed to formalise a variety of knowledge, however, whether the proposed approach is robust or sensitive to the knowledge supplied to the model has rarely been discussed. In this paper, we propose ...
Robustly Leveraging Prior Knowledge in Text Classification
2,015
http://arxiv.org/pdf/1503.00841v1
Title Robustly Leveraging Prior Knowledge Text Classification Summary Prior knowledge shown useful address many natural language processing task Many approach proposed formalise variety knowledge however whether proposed approach robust sensitive knowledge supplied model rarely discussed paper propose three regularizat...
[0.06591235846281052, 0.019557172432541847, -0.01342275645583868, 0.02875511720776558, -0.03635678067803383, 0.009935740381479263, -0.010751553811132908, 0.050245966762304306, 0.07674902677536011, -0.11466500908136368, 0.01882672682404518, 0.01053161732852459, -0.005618681199848652, 0.06086912006139755, -0.024208560585...
1,597
1,597
['Chen Liang', 'Jonathan Berant', 'Quoc Le', 'Kenneth D. Forbus', 'Ni Lao']
1612.01197v1
Extending the success of deep neural networks to natural language understanding and symbolic reasoning requires complex operations and external memory. Recent neural program induction approaches have attempted to address this problem, but are typically limited to differentiable memory, and consequently cannot scale bey...
Neural Symbolic Machines: Learning Semantic Parsers on Freebase with Weak Supervision (Short Version)
2,016
http://arxiv.org/pdf/1612.01197v1
Title Neural Symbolic Machines Learning Semantic Parsers Freebase Weak Supervision Short Version Summary Extending success deep neural network natural language understanding symbolic reasoning requires complex operation external memory Recent neural program induction approach attempted address problem typically limited...
[0.04544559493660927, 0.034564368426799774, -0.006381356623023748, 0.04750936105847359, -0.014741343446075916, 0.03231218829751015, -0.00020171802316326648, -0.022286657243967056, 0.00829470157623291, -0.03626786172389984, 0.02721855789422989, 0.03632495924830437, -0.03607761859893799, 0.04890061914920807, -0.005015165...
1,598
1,598
['Peter D. Turney']
1310.5042v1
There have been several efforts to extend distributional semantics beyond individual words, to measure the similarity of word pairs, phrases, and sentences (briefly, tuples; ordered sets of words, contiguous or noncontiguous). One way to extend beyond words is to compare two tuples using a function that combines pairwi...
Distributional semantics beyond words: Supervised learning of analogy and paraphrase
2,013
http://arxiv.org/pdf/1310.5042v1
Title Distributional semantics beyond word Supervised learning analogy paraphrase Summary several effort extend distributional semantics beyond individual word measure similarity word pair phrase sentence briefly tuples ordered set word contiguous noncontiguous One way extend beyond word compare two tuples using functi...
[0.030911607667803764, 0.06530812382698059, -0.006089201662689447, 0.04498479142785072, -0.03866848349571228, 0.028689060360193253, -0.0251599233597517, 0.007855416275560856, -0.07131210714578629, -0.049221884459257126, -0.02839106321334839, 0.010163200087845325, 0.029770033434033394, -0.018168872222304344, 0.040715612...
1,599
1,599
['Yanchuan Sim', 'Bryan Routledge', 'Noah A. Smith']
1409.7985v5
We explore the idea that authoring a piece of text is an act of maximizing one's expected utility. To make this idea concrete, we consider the societally important decisions of the Supreme Court of the United States. Extensive past work in quantitative political science provides a framework for empirically modeling the...
The Utility of Text: The Case of Amicus Briefs and the Supreme Court
2,014
http://arxiv.org/pdf/1409.7985v5
Title Utility Text Case Amicus Briefs Supreme Court Summary explore idea authoring piece text act maximizing one expected utility make idea concrete consider societally important decision Supreme Court United States Extensive past work quantitative political science provides framework empirically modeling decision just...
[0.011600423604249954, 0.07586898654699326, 0.008113550022244453, -0.02871793508529663, -0.10470691323280334, 0.03715973347425461, 0.023623621091246605, 0.020106829702854156, 0.008017083629965782, -0.028662610799074173, 0.08959028124809265, 0.01849805936217308, -0.009286652319133282, 0.028911922127008438, -0.0363101251...