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
400
400
['Volker Fischer', 'Mummadi Chaithanya Kumar', 'Jan Hendrik Metzen', 'Thomas Brox']
1703.01101v1
Machine learning methods in general and Deep Neural Networks in particular have shown to be vulnerable to adversarial perturbations. So far this phenomenon has mainly been studied in the context of whole-image classification. In this contribution, we analyse how adversarial perturbations can affect the task of semantic...
Adversarial Examples for Semantic Image Segmentation
2,017
http://arxiv.org/pdf/1703.01101v1
Title Adversarial Examples Semantic Image Segmentation Summary Machine learning method general Deep Neural Networks particular shown vulnerable adversarial perturbation far phenomenon mainly studied context wholeimage classification contribution analyse adversarial perturbation affect task semantic segmentation show ex...
[0.01410507783293724, 0.02849755436182022, -0.03793328255414963, 0.0631236806511879, -0.018297743052244186, -0.01985902152955532, 0.026824623346328735, -0.006789030507206917, -0.04114808514714241, -0.03241625055670738, 0.005177593789994717, 0.056785956025123596, 0.01314525492489338, 0.0018849825719371438, 0.06411287933...
401
401
['Wieland Brendel', 'Jonas Rauber', 'Matthias Bethge']
1712.04248v2
Many machine learning algorithms are vulnerable to almost imperceptible perturbations of their inputs. So far it was unclear how much risk adversarial perturbations carry for the safety of real-world machine learning applications because most methods used to generate such perturbations rely either on detailed model inf...
Decision-Based Adversarial Attacks: Reliable Attacks Against Black-Box Machine Learning Models
2,017
http://arxiv.org/pdf/1712.04248v2
Title DecisionBased Adversarial Attacks Reliable Attacks BlackBox Machine Learning Models Summary Many machine learning algorithm vulnerable almost imperceptible perturbation input far unclear much risk adversarial perturbation carry safety realworld machine learning application method used generate perturbation rely e...
[-0.010133789852261543, 0.041026026010513306, -0.024267854169011116, 0.025550389662384987, -0.023821262642741203, -0.0290058683604002, 0.07358074933290482, -0.026392171159386635, -0.04097146540880203, -0.04666865989565849, 0.029319528490304947, 0.0581333264708519, -0.00835547223687172, 0.06942620128393173, 0.0455332398...
402
402
['Abien Fred Agarap', 'Francis John Hill Pepito']
1801.00318v1
Effective and efficient mitigation of malware is a long-time endeavor in the information security community. The development of an anti-malware system that can counteract an unknown malware is a prolific activity that may benefit several sectors. We envision an intelligent anti-malware system that utilizes the power of...
Towards Building an Intelligent Anti-Malware System: A Deep Learning Approach using Support Vector Machine (SVM) for Malware Classification
2,017
http://arxiv.org/pdf/1801.00318v1
Title Towards Building Intelligent AntiMalware System Deep Learning Approach using Support Vector Machine SVM Malware Classification Summary Effective efficient mitigation malware longtime endeavor information security community development antimalware system counteract unknown malware prolific activity may benefit sev...
[0.042566921561956406, -0.009155770763754845, -0.037983544170856476, 0.01940065436065197, -0.02579900622367859, -0.03528895601630211, 0.03765113279223442, -0.016161033883690834, 0.017611956223845482, -0.012544902972877026, 0.05626830458641052, 0.010617326945066452, -0.003522921120747924, 0.10034801810979843, 0.05641409...
403
403
['Despoina Christou']
1604.01272v1
Feature extraction has gained increasing attention in the field of machine learning, as in order to detect patterns, extract information, or predict future observations from big data, the urge of informative features is crucial. The process of extracting features is highly linked to dimensionality reduction as it impli...
Feature extraction using Latent Dirichlet Allocation and Neural Networks: A case study on movie synopses
2,016
http://arxiv.org/pdf/1604.01272v1
Title Feature extraction using Latent Dirichlet Allocation Neural Networks case study movie synopsis Summary Feature extraction gained increasing attention field machine learning order detect pattern extract information predict future observation big data urge informative feature crucial process extracting feature high...
[0.05597390606999397, 0.044753823429346085, 0.003626729128882289, 0.05251197889447212, -0.031103627756237984, 0.002740079304203391, 0.05280204117298126, 0.013157839886844158, -0.015904661267995834, -0.03891849145293236, -0.01653800532221794, 0.008867749944329262, 0.008766223676502705, 0.04222093150019646, -0.0035424630...
404
404
['Iulian Vlad Serban', 'Ryan Lowe', 'Peter Henderson', 'Laurent Charlin', 'Joelle Pineau']
1512.05742v3
During the past decade, several areas of speech and language understanding have witnessed substantial breakthroughs from the use of data-driven models. In the area of dialogue systems, the trend is less obvious, and most practical systems are still built through significant engineering and expert knowledge. Nevertheles...
A Survey of Available Corpora for Building Data-Driven Dialogue Systems
2,015
http://arxiv.org/pdf/1512.05742v3
Title Survey Available Corpora Building DataDriven Dialogue Systems Summary past decade several area speech language understanding witnessed substantial breakthrough use datadriven model area dialogue system trend le obvious practical system still built significant engineering expert knowledge Nevertheless several rece...
[0.03697967529296875, 0.03675778582692146, -0.020406929776072502, 0.07374028116464615, 0.0004538491484709084, -0.0036602874752134085, 0.0184808149933815, -0.023244986310601234, -0.03441104665398598, -0.05692259967327118, -0.04438852146267891, -0.000322436768328771, 0.023757394403219223, 0.037595484405756, 0.01447482965...
405
405
['Shaohua Li', 'Tat-Seng Chua', 'Jun Zhu', 'Chunyan Miao']
1606.02979v2
Word embedding maps words into a low-dimensional continuous embedding space by exploiting the local word collocation patterns in a small context window. On the other hand, topic modeling maps documents onto a low-dimensional topic space, by utilizing the global word collocation patterns in the same document. These two ...
Generative Topic Embedding: a Continuous Representation of Documents (Extended Version with Proofs)
2,016
http://arxiv.org/pdf/1606.02979v2
Title Generative Topic Embedding Continuous Representation Documents Extended Version Proofs Summary Word embedding map word lowdimensional continuous embedding space exploiting local word collocation pattern small context window hand topic modeling map document onto lowdimensional topic space utilizing global word col...
[0.012248310260474682, 0.016735142096877098, -0.0054135615937411785, 0.018747160211205482, -0.011101397685706615, 0.009182619862258434, 0.0185473021119833, -0.005352766718715429, -0.03516506403684616, -0.05392949655652046, 0.012300546281039715, 0.008827975951135159, 0.011603852733969688, 0.0505647175014019, 0.026435902...
406
406
['Maxim Rabinovich', 'Dan Klein']
1704.07751v1
As entity type systems become richer and more fine-grained, we expect the number of types assigned to a given entity to increase. However, most fine-grained typing work has focused on datasets that exhibit a low degree of type multiplicity. In this paper, we consider the high-multiplicity regime inherent in data source...
Fine-Grained Entity Typing with High-Multiplicity Assignments
2,017
http://arxiv.org/pdf/1704.07751v1
Title FineGrained Entity Typing HighMultiplicity Assignments Summary entity type system become richer finegrained expect number type assigned given entity increase However finegrained typing work focused datasets exhibit low degree type multiplicity paper consider highmultiplicity regime inherent data source Wikipedia ...
[0.0005161462468095124, 0.026029350236058235, -0.01804785244166851, 0.00019324070308357477, 0.006674001459032297, 0.028266053646802902, 0.03170699253678322, 0.020208759233355522, -0.02069169096648693, -0.06634200364351273, 0.010585351847112179, -0.021944336593151093, 0.018650393933057785, 0.0321228913962841, 0.05646652...
407
407
['Mateusz Malinowski', 'Mario Fritz']
1410.8027v3
As language and visual understanding by machines progresses rapidly, we are observing an increasing interest in holistic architectures that tightly interlink both modalities in a joint learning and inference process. This trend has allowed the community to progress towards more challenging and open tasks and refueled t...
Towards a Visual Turing Challenge
2,014
http://arxiv.org/pdf/1410.8027v3
Title Towards Visual Turing Challenge Summary language visual understanding machine progress rapidly observing increasing interest holistic architecture tightly interlink modality joint learning inference process trend allowed community progress towards challenging open task refueled hope achieving old AI dream buildin...
[0.0014101818669587374, 0.026177240535616875, -0.008235828951001167, 0.003800810780376196, -0.04008960723876953, -0.017938468605279922, 0.017642751336097717, 0.016304917633533478, -0.0054808142594993114, 0.02546154521405697, 0.004586879629641771, 0.07738128304481506, 0.013647174462676048, 0.10973093658685684, 0.0121642...
408
408
['Giovanni Saponaro', 'Lorenzo Jamone', 'Alexandre Bernardino', 'Giampiero Salvi']
1711.09055v1
A growing field in robotics and Artificial Intelligence (AI) research is human-robot collaboration, whose target is to enable effective teamwork between humans and robots. However, in many situations human teams are still superior to human-robot teams, primarily because human teams can easily agree on a common goal wit...
Interactive Robot Learning of Gestures, Language and Affordances
2,017
http://arxiv.org/pdf/1711.09055v1
Title Interactive Robot Learning Gestures Language Affordances Summary growing field robotics Artificial Intelligence AI research humanrobot collaboration whose target enable effective teamwork human robot However many situation human team still superior humanrobot team primarily human team easily agree common goal lan...
[0.027822798117995262, 0.018308086320757866, 0.004649443551898003, 0.0024462572764605284, -0.014528930187225342, 0.050836119800806046, -0.0033683222718536854, 0.003362779738381505, 0.04727567732334137, -0.043453603982925415, -0.026441389694809914, 0.01374482549726963, 0.031093403697013855, 0.03950246050953865, -0.00082...
409
409
['Abhinav Gupta', 'Yajie Miao', 'Leonardo Neves', 'Florian Metze']
1712.00489v1
Automatic transcriptions of consumer-generated multi-media content such as "Youtube" videos still exhibit high word error rates. Such data typically occupies a very broad domain, has been recorded in challenging conditions, with cheap hardware and a focus on the visual modality, and may have been post-processed or edit...
Visual Features for Context-Aware Speech Recognition
2,017
http://arxiv.org/pdf/1712.00489v1
Title Visual Features ContextAware Speech Recognition Summary Automatic transcription consumergenerated multimedia content Youtube video still exhibit high word error rate data typically occupies broad domain recorded challenging condition cheap hardware focus visual modality may postprocessed edited paper extend earli...
[0.0262033361941576, 0.04302217438817024, 0.01628383994102478, 0.030029460787773132, -0.01211551297456026, 0.0027189040556550026, 0.026020489633083344, 0.02941032312810421, -0.07764282822608948, -0.09758231788873672, -0.01810879074037075, 0.020536627620458603, 0.04438067972660065, 0.04513340815901756, 0.002862241351976...
410
410
['Mircea Mironenco', 'Dana Kianfar', 'Ke Tran', 'Evangelos Kanoulas', 'Efstratios Gavves']
1712.01329v1
In this work we propose a blackbox intervention method for visual dialog models, with the aim of assessing the contribution of individual linguistic or visual components. Concretely, we conduct structured or randomized interventions that aim to impair an individual component of the model, and observe changes in task pe...
Examining Cooperation in Visual Dialog Models
2,017
http://arxiv.org/pdf/1712.01329v1
Title Examining Cooperation Visual Dialog Models Summary work propose blackbox intervention method visual dialog model aim assessing contribution individual linguistic visual component Concretely conduct structured randomized intervention aim impair individual component model observe change task performance reproduce s...
[0.08261365443468094, -0.001558822812512517, -0.053738925606012344, 0.013611875474452972, -0.054988399147987366, 0.022216124460101128, 0.04457809031009674, 0.045046523213386536, 0.017086386680603027, -0.04278905317187309, -0.01808173768222332, -0.020458683371543884, 0.0247054323554039, 0.07439517229795456, -0.000651815...
411
411
['Cheng-Yang Fu', 'Joon Lee', 'Mohit Bansal', 'Alexander C. Berg']
1707.08559v1
Sports channel video portals offer an exciting domain for research on multimodal, multilingual analysis. We present methods addressing the problem of automatic video highlight prediction based on joint visual features and textual analysis of the real-world audience discourse with complex slang, in both English and trad...
Video Highlight Prediction Using Audience Chat Reactions
2,017
http://arxiv.org/pdf/1707.08559v1
Title Video Highlight Prediction Using Audience Chat Reactions Summary Sports channel video portal offer exciting domain research multimodal multilingual analysis present method addressing problem automatic video highlight prediction based joint visual feature textual analysis realworld audience discourse complex slang...
[0.03348467871546745, 0.007148250471800566, -0.01630496047437191, 0.0561295747756958, 0.026042751967906952, 0.004499316215515137, -0.009511063806712627, 0.03231172263622284, -0.019246330484747887, -0.07985155284404755, 0.009361283853650093, -0.05005268380045891, -0.02249239571392536, 0.08811645209789276, 0.024427358061...
412
412
['Dmitriy Serdyuk', 'Kartik Audhkhasi', 'Philémon Brakel', 'Bhuvana Ramabhadran', 'Samuel Thomas', 'Yoshua Bengio']
1612.01928v1
Modern automatic speech recognition (ASR) systems need to be robust under acoustic variability arising from environmental, speaker, channel, and recording conditions. Ensuring such robustness to variability is a challenge in modern day neural network-based ASR systems, especially when all types of variability are not s...
Invariant Representations for Noisy Speech Recognition
2,016
http://arxiv.org/pdf/1612.01928v1
Title Invariant Representations Noisy Speech Recognition Summary Modern automatic speech recognition ASR system need robust acoustic variability arising environmental speaker channel recording condition Ensuring robustness variability challenge modern day neural networkbased ASR system especially type variability seen ...
[-0.0066824546083807945, 0.041211169213056564, -0.0013693394139409065, 0.020190533250570297, 0.03499221429228783, -0.011152408085763454, 0.05345822498202324, -0.023711681365966797, -0.03150981664657593, -0.020177843049168587, -0.06726479530334473, -0.011691833846271038, 0.016218282282352448, 0.02350870706140995, 0.0457...
413
413
['Kalin Stefanov', 'Jonas Beskow', 'Giampiero Salvi']
1711.08992v1
This paper presents a self-supervised method for detecting the active speaker in a multi-person spoken interaction scenario. We argue that this capability is a fundamental prerequisite for any artificial cognitive system attempting to acquire language in social settings. Our methods are able to detect an arbitrary numb...
Self-Supervised Vision-Based Detection of the Active Speaker as a Prerequisite for Socially-Aware Language Acquisition
2,017
http://arxiv.org/pdf/1711.08992v1
Title SelfSupervised VisionBased Detection Active Speaker Prerequisite SociallyAware Language Acquisition Summary paper present selfsupervised method detecting active speaker multiperson spoken interaction scenario argue capability fundamental prerequisite artificial cognitive system attempting acquire language social ...
[0.026632023975253105, -0.03483507037162781, -0.027896825224161148, 0.05321666598320007, -0.004127347841858864, 0.010405980981886387, 0.005493324249982834, -0.02799266017973423, 0.050446610897779465, -0.02850143425166607, -0.015240682289004326, 0.018956303596496582, 0.039852406829595566, 0.01963954232633114, 0.02493328...
414
414
['Ângelo Cardoso', 'Fabio Daolio', 'Saúl Vargas']
1803.07679v1
In this paper, we describe a solution to tackle a common set of challenges in e-commerce, which arise from the fact that new products are continually being added to the catalogue. The challenges involve properly personalising the customer experience, forecasting demand and planning the product range. We argue that the ...
Product Characterisation towards Personalisation: Learning Attributes from Unstructured Data to Recommend Fashion Products
2,018
http://arxiv.org/pdf/1803.07679v1
Title Product Characterisation towards Personalisation Learning Attributes Unstructured Data Recommend Fashion Products Summary paper describe solution tackle common set challenge ecommerce arise fact new product continually added catalogue challenge involve properly personalising customer experience forecasting demand...
[0.014864067547023296, 0.0035405221860855818, -0.019473256543278694, 0.013838011771440506, 0.015247411094605923, -0.018700018525123596, 0.04886321723461151, -0.003940575290471315, -0.02153913304209709, -0.03618629649281502, -0.019771840423345566, 0.027456816285848618, 0.003578206291422248, 0.10362014919519424, -0.03508...
415
415
['Pierre-Yves Oudeyer']
cs/0502086v1
The speech code is a vehicle of language: it defines a set of forms used by a community to carry information. Such a code is necessary to support the linguistic interactions that allow humans to communicate. How then may a speech code be formed prior to the existence of linguistic interactions? Moreover, the human spee...
The Self-Organization of Speech Sounds
2,005
http://arxiv.org/pdf/cs/0502086v1
Title SelfOrganization Speech Sounds Summary speech code vehicle language defines set form used community carry information code necessary support linguistic interaction allow human communicate may speech code formed prior existence linguistic interaction Moreover human speech code discrete compositional shared individ...
[0.004522000905126333, 0.037893615663051605, -0.02303241938352585, -0.00526499655097723, 0.0034601346123963594, -0.005060199648141861, 0.0422058142721653, 0.004528329242020845, -0.0005553242517635226, -0.027725689113140106, -0.04336715489625931, 0.05382722243666649, 0.034057144075632095, 0.04156311973929405, 0.05471818...
416
416
['David M. W. Powers']
1503.06410v1
The F-measure or F-score is one of the most commonly used single number measures in Information Retrieval, Natural Language Processing and Machine Learning, but it is based on a mistake, and the flawed assumptions render it unsuitable for use in most contexts! Fortunately, there are better alternatives.
What the F-measure doesn't measure: Features, Flaws, Fallacies and Fixes
2,015
http://arxiv.org/pdf/1503.06410v1
Title Fmeasure doesnt measure Features Flaws Fallacies Fixes Summary Fmeasure Fscore one commonly used single number measure Information Retrieval Natural Language Processing Machine Learning based mistake flawed assumption render unsuitable use context Fortunately better alternative Authors 0 Ahmed Osman Wojciech Same...
[0.02853039652109146, 0.03242217376828194, -0.027932828292250633, -0.008576486259698868, -0.023038577288389206, 0.014914223924279213, 0.04338524490594864, 0.003689747303724289, -0.0029782666824758053, -0.015994569286704063, 0.022381436079740524, -0.04872889071702957, 0.0673162192106247, 0.061363864690065384, -0.0245049...
417
417
['Byron Knoll', 'Nando de Freitas']
1108.3298v1
PAQ8 is an open source lossless data compression algorithm that currently achieves the best compression rates on many benchmarks. This report presents a detailed description of PAQ8 from a statistical machine learning perspective. It shows that it is possible to understand some of the modules of PAQ8 and use this under...
A Machine Learning Perspective on Predictive Coding with PAQ
2,011
http://arxiv.org/pdf/1108.3298v1
Title Machine Learning Perspective Predictive Coding PAQ Summary PAQ8 open source lossless data compression algorithm currently achieves best compression rate many benchmark report present detailed description PAQ8 statistical machine learning perspective show possible understand module PAQ8 use understanding improve m...
[-0.023787135258316994, 0.0296892449259758, -0.012251048348844051, 0.03194970637559891, -0.00995184201747179, 0.002444350393489003, -0.03844984620809555, 0.062469519674777985, -0.07234214246273041, -0.0002977386175189167, 0.019992707297205925, 0.013499472290277481, 0.03565507009625435, 0.08334913104772568, 0.0083262789...
418
418
['Hector Allende', 'Emanuele Frandi', 'Ricardo Nanculef', 'Claudio Sartori']
1304.1014v2
Recently, there has been a renewed interest in the machine learning community for variants of a sparse greedy approximation procedure for concave optimization known as {the Frank-Wolfe (FW) method}. In particular, this procedure has been successfully applied to train large-scale instances of non-linear Support Vector M...
A Novel Frank-Wolfe Algorithm. Analysis and Applications to Large-Scale SVM Training
2,013
http://arxiv.org/pdf/1304.1014v2
Title Novel FrankWolfe Algorithm Analysis Applications LargeScale SVM Training Summary Recently renewed interest machine learning community variant sparse greedy approximation procedure concave optimization known FrankWolfe FW method particular procedure successfully applied train largescale instance nonlinear Support ...
[0.00035654165549203753, -0.011363559402525425, 0.0011729124234989285, 0.04786599427461624, 0.008152477443218231, 0.0029361483175307512, -0.0018268664134666324, 0.023853057995438576, -0.0084852809086442, -0.02327011339366436, 0.007954495958983898, -0.032711416482925415, 0.02722095511853695, -0.009915714152157307, 0.001...
419
419
['Yanwei Fu', 'Leonid Sigal']
1604.07093v1
Despite significant progress in object categorization, in recent years, a number of important challenges remain, mainly, ability to learn from limited labeled data and ability to recognize object classes within large, potentially open, set of labels. Zero-shot learning is one way of addressing these challenges, but it ...
Semi-supervised Vocabulary-informed Learning
2,016
http://arxiv.org/pdf/1604.07093v1
Title Semisupervised Vocabularyinformed Learning Summary Despite significant progress object categorization recent year number important challenge remain mainly ability learn limited labeled data ability recognize object class within large potentially open set label Zeroshot learning one way addressing challenge shown ...
[-0.019598769024014473, 0.014297908172011375, 0.0270952470600605, 0.08113818615674973, 0.007989886216819286, 0.011461552232503891, 0.046187061816453934, 0.01979505456984043, -0.005289588123559952, -0.0718226432800293, -0.030910111963748932, 0.01094746496528387, -0.07155538350343704, 0.03528866171836853, -0.010574105195...
420
420
['Adarsh Prasad', 'Stefanie Jegelka', 'Dhruv Batra']
1411.1752v1
To cope with the high level of ambiguity faced in domains such as Computer Vision or Natural Language processing, robust prediction methods often search for a diverse set of high-quality candidate solutions or proposals. In structured prediction problems, this becomes a daunting task, as the solution space (image label...
Submodular meets Structured: Finding Diverse Subsets in Exponentially-Large Structured Item Sets
2,014
http://arxiv.org/pdf/1411.1752v1
Title Submodular meet Structured Finding Diverse Subsets ExponentiallyLarge Structured Item Sets Summary cope high level ambiguity faced domain Computer Vision Natural Language processing robust prediction method often search diverse set highquality candidate solution proposal structured prediction problem becomes daun...
[0.04298301786184311, 0.014171047136187553, -0.023479577153921127, 0.036026570945978165, -0.012289566919207573, -0.012425921857357025, 0.007665161043405533, 0.02899244986474514, 0.035471104085445404, -0.07495713979005814, 0.005614526104182005, -0.035461537539958954, 0.016258588060736656, 0.08860262483358383, 0.00868114...
421
421
['Hao Wang', 'Xiaodan Liang', 'Hao Zhang', 'Dit-Yan Yeung', 'Eric P. Xing']
1703.07255v2
Many problems in image processing and computer vision (e.g. colorization, style transfer) can be posed as 'manipulating' an input image into a corresponding output image given a user-specified guiding signal. A holy-grail solution towards generic image manipulation should be able to efficiently alter an input image wit...
ZM-Net: Real-time Zero-shot Image Manipulation Network
2,017
http://arxiv.org/pdf/1703.07255v2
Title ZMNet Realtime Zeroshot Image Manipulation Network Summary Many problem image processing computer vision eg colorization style transfer posed manipulating input image corresponding output image given userspecified guiding signal holygrail solution towards generic image manipulation able efficiently alter input im...
[-0.00017065303109120578, 0.02034536749124527, 0.028135545551776886, 0.03961324319243431, -0.01038052886724472, -0.014034738764166832, 0.04388752579689026, 0.024866612628102303, -0.10656560212373734, -0.01895078271627426, -0.053714558482170105, 0.03644692897796631, -0.023473083972930908, 0.024208905175328255, 0.0166290...
422
422
['Arnab Ghosh', 'Viveka Kulharia', 'Vinay Namboodiri', 'Philip H. S. Torr', 'Puneet K. Dokania']
1704.02906v2
We propose an intuitive generalization to the Generative Adversarial Networks (GANs) and its conditional variants to address the well known mode collapse problem. Firstly, we propose a multi-agent GAN architecture incorporating multiple generators and one discriminator. Secondly, to enforce different generators to capt...
Multi-Agent Diverse Generative Adversarial Networks
2,017
http://arxiv.org/pdf/1704.02906v2
Title MultiAgent Diverse Generative Adversarial Networks Summary propose intuitive generalization Generative Adversarial Networks GANs conditional variant address well known mode collapse problem Firstly propose multiagent GAN architecture incorporating multiple generator one discriminator Secondly enforce different ge...
[-0.006788115948438644, 0.08157891780138016, -0.013384711928665638, 0.011438085697591305, 0.008202669210731983, 0.0033263543155044317, 0.07081679254770279, -0.004084078595042229, -0.036281030625104904, 0.029036708176136017, -0.0325741320848465, -0.011441606096923351, -0.03196254372596741, 0.018260378390550613, 0.076945...
423
423
['Jae Hyun Lim', 'Jong Chul Ye']
1705.02894v2
Generative Adversarial Nets (GANs) represent an important milestone for effective generative models, which has inspired numerous variants seemingly different from each other. One of the main contributions of this paper is to reveal a unified geometric structure in GAN and its variants. Specifically, we show that the ad...
Geometric GAN
2,017
http://arxiv.org/pdf/1705.02894v2
Title Geometric GAN Summary Generative Adversarial Nets GANs represent important milestone effective generative model inspired numerous variant seemingly different One main contribution paper reveal unified geometric structure GAN variant Specifically show adversarial generative model training decomposed three geometri...
[0.02296755649149418, 0.03869876265525818, -0.02602660283446312, 0.03212890401482582, 0.008779403753578663, 0.005982937291264534, 0.03460725024342537, -0.03673403337597847, -0.03673959895968437, 0.0370747447013855, -0.0005408736760728061, 0.02223697490990162, -0.02752140536904335, 0.018071528524160385, 0.09001615643501...
424
424
['Disha Shrivastava', 'Santanu Chaudhury', 'Dr. Jayadeva']
1708.05840v1
Training deep networks is expensive and time-consuming with the training period increasing with data size and growth in model parameters. In this paper, we provide a framework for distributed training of deep networks over a cluster of CPUs in Apache Spark. The framework implements both Data Parallelism and Model Paral...
A Data and Model-Parallel, Distributed and Scalable Framework for Training of Deep Networks in Apache Spark
2,017
http://arxiv.org/pdf/1708.05840v1
Title Data ModelParallel Distributed Scalable Framework Training Deep Networks Apache Spark Summary Training deep network expensive timeconsuming training period increasing data size growth model parameter paper provide framework distributed training deep network cluster CPUs Apache Spark framework implement Data Paral...
[-0.0018520618323236704, 0.01606283150613308, -0.014184780418872833, 0.07739543914794922, 0.008176264353096485, 0.006769976112991571, 0.06531637161970139, -0.003001242643222213, 0.004967254586517811, -0.012263813987374306, -0.005408227909356356, 0.029699349775910378, -0.021033965051174164, 0.06446307897567749, -0.00516...
425
425
['Sebastian Lapuschkin', 'Alexander Binder', 'Klaus-Robert Müller', 'Wojciech Samek']
1708.07689v1
Recently, deep neural networks have demonstrated excellent performances in recognizing the age and gender on human face images. However, these models were applied in a black-box manner with no information provided about which facial features are actually used for prediction and how these features depend on image prepro...
Understanding and Comparing Deep Neural Networks for Age and Gender Classification
2,017
http://arxiv.org/pdf/1708.07689v1
Title Understanding Comparing Deep Neural Networks Age Gender Classification Summary Recently deep neural network demonstrated excellent performance recognizing age gender human face image However model applied blackbox manner information provided facial feature actually used prediction feature depend image preprocessi...
[0.026481812819838524, 0.03845563903450966, -0.015925094485282898, 0.024296240881085396, 0.008811231702566147, 0.03944318741559982, 0.0840742439031601, -0.01714557781815529, -0.006307459436357021, 0.015625886619091034, 0.0006790544139221311, -0.06746894866228104, 0.04517519101500511, 0.0567784383893013, 0.0085363583639...
426
426
['Simon S. Du', 'Jason D. Lee', 'Yuandong Tian']
1709.06129v2
We analyze the convergence of (stochastic) gradient descent algorithm for learning a convolutional filter with Rectified Linear Unit (ReLU) activation function. Our analysis does not rely on any specific form of the input distribution and our proofs only use the definition of ReLU, in contrast with previous works that ...
When is a Convolutional Filter Easy To Learn?
2,017
http://arxiv.org/pdf/1709.06129v2
Title Convolutional Filter Easy Learn Summary analyze convergence stochastic gradient descent algorithm learning convolutional filter Rectified Linear Unit ReLU activation function analysis rely specific form input distribution proof use definition ReLU contrast previous work restricted standard Gaussian input show sto...
[-0.0328933484852314, 0.0046985335648059845, 0.0005247725057415664, 0.07850399613380432, 0.01807556487619877, -0.06688686460256577, 0.01794382743537426, 0.021067945286631584, -0.027484670281410217, 0.01705227978527546, 0.004229845944792032, 0.009164389222860336, -0.004456693306565285, 0.06976305693387985, -0.0015496979...
427
427
['Jianqiao Wangni', 'Dahua Lin']
1711.02857v1
Sparsity inducing regularization is an important part for learning over-complete visual representations. Despite the popularity of $\ell_1$ regularization, in this paper, we investigate the usage of non-convex regularizations in this problem. Our contribution consists of three parts. First, we propose the leaky capped ...
Learning Sparse Visual Representations with Leaky Capped Norm Regularizers
2,017
http://arxiv.org/pdf/1711.02857v1
Title Learning Sparse Visual Representations Leaky Capped Norm Regularizers Summary Sparsity inducing regularization important part learning overcomplete visual representation Despite popularity ell1 regularization paper investigate usage nonconvex regularization problem contribution consists three part First propose l...
[-0.008767512626945972, 0.04537312686443329, 0.006528915371745825, 0.05196036025881767, 0.010393067263066769, -0.022254910320043564, -0.014145621098577976, -0.014949527569115162, 0.0015199027257040143, 0.03983422368764877, -0.002558992011472583, 0.005046372767537832, 0.012634659186005592, 0.06136992573738098, 0.0481484...
428
428
['Pierre Stock', 'Moustapha Cisse']
1711.11443v1
ConvNets and Imagenet have driven the recent success of deep learning for image classification. However, the marked slowdown in performance improvement, the recent studies on the lack of robustness of neural networks to adversarial examples and their tendency to exhibit undesirable biases (e.g racial biases) questioned...
ConvNets and ImageNet Beyond Accuracy: Explanations, Bias Detection, Adversarial Examples and Model Criticism
2,017
http://arxiv.org/pdf/1711.11443v1
Title ConvNets ImageNet Beyond Accuracy Explanations Bias Detection Adversarial Examples Model Criticism Summary ConvNets Imagenet driven recent success deep learning image classification However marked slowdown performance improvement recent study lack robustness neural network adversarial example tendency exhibit und...
[0.013222314417362213, 0.05636214837431908, -0.03590932860970497, 0.04753301665186882, 0.016025608405470848, -0.0021171909756958485, 0.0635276809334755, -0.0063713788986206055, -0.00709938257932663, -0.0005602426826953888, -0.012481095269322395, 0.024519531056284904, -0.0032368318643420935, 0.02208569459617138, 0.01956...
429
429
['Simon S. Du', 'Jason D. Lee', 'Yuandong Tian', 'Barnabas Poczos', 'Aarti Singh']
1712.00779v1
We consider the problem of learning a one-hidden-layer neural network with non-overlapping convolutional layer and ReLU activation function, i.e., $f(\mathbf{Z}; \mathbf{w}, \mathbf{a}) = \sum_j a_j\sigma(\mathbf{w}^\top\mathbf{Z}_j)$, in which both the convolutional weights $\mathbf{w}$ and the output weights $\mathbf...
Gradient Descent Learns One-hidden-layer CNN: Don't be Afraid of Spurious Local Minima
2,017
http://arxiv.org/pdf/1712.00779v1
Title Gradient Descent Learns Onehiddenlayer CNN Dont Afraid Spurious Local Minima Summary consider problem learning onehiddenlayer neural network nonoverlapping convolutional layer ReLU activation function ie fmathbfZ mathbfw mathbfa sumj ajsigmamathbfwtopmathbfZj convolutional weight mathbfw output weight mathbfa par...
[-0.015815701335668564, 0.0034348166082054377, 0.004034135490655899, 0.0038254817482084036, 0.00422449316829443, -0.04407276213169098, -0.015035745687782764, 0.004122546873986721, -0.06227981299161911, 0.05880111828446388, 0.027219809591770172, 0.01928306370973587, 0.016228588297963142, -0.0013058217009529471, -0.00276...
430
430
['Deepak Pathak', 'Pulkit Agrawal', 'Alexei A. Efros', 'Trevor Darrell']
1705.05363v1
In many real-world scenarios, rewards extrinsic to the agent are extremely sparse, or absent altogether. In such cases, curiosity can serve as an intrinsic reward signal to enable the agent to explore its environment and learn skills that might be useful later in its life. We formulate curiosity as the error in an agen...
Curiosity-driven Exploration by Self-supervised Prediction
2,017
http://arxiv.org/pdf/1705.05363v1
Title Curiositydriven Exploration Selfsupervised Prediction Summary many realworld scenario reward extrinsic agent extremely sparse absent altogether case curiosity serve intrinsic reward signal enable agent explore environment learn skill might useful later life formulate curiosity error agent ability predict conseque...
[0.019236771389842033, 0.03240903839468956, -0.024170326068997383, 0.008853362873196602, 0.015394032001495361, -0.02874867245554924, 0.008909371681511402, -0.008354834280908108, -0.021054735407233238, 0.03710471838712692, -0.008091621100902557, 0.0486241839826107, -0.06268144398927689, 0.08134281635284424, 0.0042772707...
431
431
['Moustapha Cisse', 'Yossi Adi', 'Natalia Neverova', 'Joseph Keshet']
1707.05373v1
Generating adversarial examples is a critical step for evaluating and improving the robustness of learning machines. So far, most existing methods only work for classification and are not designed to alter the true performance measure of the problem at hand. We introduce a novel flexible approach named Houdini for gene...
Houdini: Fooling Deep Structured Prediction Models
2,017
http://arxiv.org/pdf/1707.05373v1
Title Houdini Fooling Deep Structured Prediction Models Summary Generating adversarial example critical step evaluating improving robustness learning machine far existing method work classification designed alter true performance measure problem hand introduce novel flexible approach named Houdini generating adversaria...
[0.011369560845196247, 0.05320367589592934, -0.024191319942474365, 0.04087435454130173, -0.0031127736438065767, -0.005763102788478136, 0.016725635156035423, -0.0004394883580971509, 0.008062866516411304, -0.031695492565631866, -0.045212436467409134, 0.03004785254597664, 0.015374635346233845, 0.02745014615356922, 0.03978...
432
432
['Yanwei Fu', 'Tao Xiang', 'Yu-Gang Jiang', 'Xiangyang Xue', 'Leonid Sigal', 'Shaogang Gong']
1710.04837v1
With the recent renaissance of deep convolution neural networks, encouraging breakthroughs have been achieved on the supervised recognition tasks, where each class has sufficient training data and fully annotated training data. However, to scale the recognition to a large number of classes with few or now training samp...
Recent Advances in Zero-shot Recognition
2,017
http://arxiv.org/pdf/1710.04837v1
Title Recent Advances Zeroshot Recognition Summary recent renaissance deep convolution neural network encouraging breakthrough achieved supervised recognition task class sufficient training data fully annotated training data However scale recognition large number class training sample class remains unsolved problem One...
[-0.07141465693712234, -0.005048695486038923, 0.027985263615846634, 0.04772936925292015, -0.01042885147035122, -0.005197441205382347, 0.09112165123224258, 0.05420561134815216, -0.058816999197006226, -0.053750135004520416, 0.0024923193268477917, 0.00012013706145808101, -0.0037958689499646425, 0.0007211341289803386, 0.00...
433
433
['Quynh Nguyen', 'Matthias Hein']
1710.10928v1
We analyze the expressiveness and loss surface of practical deep convolutional neural networks (CNNs) with shared weights and max pooling layers. We show that such CNNs produce linearly independent features at a "wide" layer which has more neurons than the number of training samples. This condition holds e.g. for the V...
The loss surface and expressivity of deep convolutional neural networks
2,017
http://arxiv.org/pdf/1710.10928v1
Title loss surface expressivity deep convolutional neural network Summary analyze expressiveness loss surface practical deep convolutional neural network CNNs shared weight max pooling layer show CNNs produce linearly independent feature wide layer neuron number training sample condition hold eg VGG network Furthermore...
[-0.004566425923258066, 0.00406654505059123, 0.002062950748950243, 0.0939943864941597, 0.02027236297726631, -0.02716669999063015, 0.024386368691921234, -0.003845380851998925, -0.05316748842597008, 0.019360898062586784, -0.007746878080070019, 0.041482966393232346, -0.020502062514424324, 0.039715688675642014, 0.050672046...
434
434
['Anuj Karpatne', 'William Watkins', 'Jordan Read', 'Vipin Kumar']
1710.11431v2
This paper introduces a novel framework for combining scientific knowledge of physics-based models with neural networks to advance scientific discovery. This framework, termed as physics-guided neural network (PGNN), leverages the output of physics-based model simulations along with observational features to generate p...
Physics-guided Neural Networks (PGNN): An Application in Lake Temperature Modeling
2,017
http://arxiv.org/pdf/1710.11431v2
Title Physicsguided Neural Networks PGNN Application Lake Temperature Modeling Summary paper introduces novel framework combining scientific knowledge physicsbased model neural network advance scientific discovery framework termed physicsguided neural network PGNN leverage output physicsbased model simulation along obs...
[-0.02649982087314129, -0.029062263667583466, -0.009417790919542313, 0.02430092729628086, -0.005859453696757555, -0.029969457536935806, 0.025071565061807632, -0.02109672501683235, 0.05818287283182144, -0.0058889687061309814, -0.0374637246131897, 0.025711659342050552, 0.06218906119465828, 0.03147970139980316, 0.05102292...
435
435
['Zhao Kang', 'Chong Peng', 'Qiang Cheng', 'Zenglin Xu']
1711.04258v1
Spectral clustering has found extensive use in many areas. Most traditional spectral clustering algorithms work in three separate steps: similarity graph construction; continuous labels learning; discretizing the learned labels by k-means clustering. Such common practice has two potential flaws, which may lead to sever...
Unified Spectral Clustering with Optimal Graph
2,017
http://arxiv.org/pdf/1711.04258v1
Title Unified Spectral Clustering Optimal Graph Summary Spectral clustering found extensive use many area traditional spectral clustering algorithm work three separate step similarity graph construction continuous label learning discretizing learned label kmeans clustering common practice two potential flaw may lead se...
[-0.015670178458094597, -0.024089861661195755, -0.012890690937638283, 0.03483209013938904, -0.009417342022061348, -0.02201218530535698, 0.02096923440694809, -0.0009460219880566001, 0.029986942186951637, -0.0007581560057587922, -0.035129230469465256, 0.04066723585128784, 0.016056865453720093, -0.004265537951141596, -0.0...
436
436
['David P. Helmbold', 'Philip M. Long']
1412.4736v4
Dropout is a simple but effective technique for learning in neural networks and other settings. A sound theoretical understanding of dropout is needed to determine when dropout should be applied and how to use it most effectively. In this paper we continue the exploration of dropout as a regularizer pioneered by Wager,...
On the Inductive Bias of Dropout
2,014
http://arxiv.org/pdf/1412.4736v4
Title Inductive Bias Dropout Summary Dropout simple effective technique learning neural network setting sound theoretical understanding dropout needed determine dropout applied use effectively paper continue exploration dropout regularizer pioneered Wager etal focus linear classification convex proxy misclassification ...
[-0.005859412252902985, 0.052522558718919754, -0.025943441316485405, 0.017594603821635246, 0.007508797105401754, -0.026685962453484535, 0.06674736738204956, -8.999822603072971e-05, -0.0064962864853441715, 0.02773449383676052, 0.019221490249037743, 0.07637382298707962, 0.01978726126253605, 0.025818446651101112, 0.030248...
437
437
['David P. Helmbold', 'Philip M. Long']
1602.04484v5
We analyze dropout in deep networks with rectified linear units and the quadratic loss. Our results expose surprising differences between the behavior of dropout and more traditional regularizers like weight decay. For example, on some simple data sets dropout training produces negative weights even though the output i...
Surprising properties of dropout in deep networks
2,016
http://arxiv.org/pdf/1602.04484v5
Title Surprising property dropout deep network Summary analyze dropout deep network rectified linear unit quadratic loss result expose surprising difference behavior dropout traditional regularizers like weight decay example simple data set dropout training produce negative weight even though output sum input provides ...
[-0.03788214549422264, 0.061066094785928726, -0.014872054569423199, 0.028878727927803993, 0.01833534799516201, -0.010069230571389198, 0.030231766402721405, -0.006657672580331564, -0.01702852174639702, 0.052752669900655746, 0.009050716646015644, 0.06037207692861557, 0.00027876460808329284, 0.08301247656345367, 0.0140492...
438
438
['Alireza Bagheri', 'Osvaldo Simeone', 'Bipin Rajendran']
1710.10704v3
Third-generation neural networks, or Spiking Neural Networks (SNNs), aim at harnessing the energy efficiency of spike-domain processing by building on computing elements that operate on, and exchange, spikes. In this paper, the problem of training a two-layer SNN is studied for the purpose of classification, under a Ge...
Training Probabilistic Spiking Neural Networks with First-to-spike Decoding
2,017
http://arxiv.org/pdf/1710.10704v3
Title Training Probabilistic Spiking Neural Networks Firsttospike Decoding Summary Thirdgeneration neural network Spiking Neural Networks SNNs aim harnessing energy efficiency spikedomain processing building computing element operate exchange spike paper problem training twolayer SNN studied purpose classification Gene...
[-0.039866380393505096, -0.04846152663230896, -0.022309591993689537, 0.05436500534415245, 0.01894194819033146, 0.0014488634187728167, 0.01863725483417511, -0.0003226301050744951, -0.001143521280027926, -0.008925831876695156, -0.03076167032122612, 0.022754700854420662, 0.022677382454276085, 0.08209702372550964, 0.028869...
439
439
['Qiang Li', 'Yan He', 'Jing-ping Jiang']
0812.0743v2
Enormous successes have been made by quantum algorithms during the last decade. In this paper, we combine the quantum game with the problem of data clustering, and then develop a quantum-game-based clustering algorithm, in which data points in a dataset are considered as players who can make decisions and implement qua...
A Novel Clustering Algorithm Based on Quantum Games
2,008
http://arxiv.org/pdf/0812.0743v2
Title Novel Clustering Algorithm Based Quantum Games Summary Enormous success made quantum algorithm last decade paper combine quantum game problem data clustering develop quantumgamebased clustering algorithm data point dataset considered player make decision implement quantum strategy quantum game round quantum game ...
[-0.02627340331673622, 0.014408312737941742, -0.07075487822294235, 0.05268136039376259, -0.03348598629236221, -0.019654547795653343, -0.02003326825797558, 0.035398319363594055, 0.042569659650325775, 0.011598685756325722, -0.029186438769102097, -0.009110826067626476, -0.011148972436785698, -0.0004399906028993428, 0.0229...
440
440
['Andrew M. Saxe', 'James L. McClelland', 'Surya Ganguli']
1312.6120v3
Despite the widespread practical success of deep learning methods, our theoretical understanding of the dynamics of learning in deep neural networks remains quite sparse. We attempt to bridge the gap between the theory and practice of deep learning by systematically analyzing learning dynamics for the restricted case o...
Exact solutions to the nonlinear dynamics of learning in deep linear neural networks
2,013
http://arxiv.org/pdf/1312.6120v3
Title Exact solution nonlinear dynamic learning deep linear neural network Summary Despite widespread practical success deep learning method theoretical understanding dynamic learning deep neural network remains quite sparse attempt bridge gap theory practice deep learning systematically analyzing learning dynamic rest...
[-0.014215430244803429, 0.001987025374546647, -0.04314752668142319, 0.006781932432204485, 0.023280799388885498, -0.014689714647829533, -0.00482304347679019, -0.021557992324233055, -0.043977826833724976, 0.03281954675912857, 0.010095802135765553, 0.03388497233390808, -0.020579464733600616, 0.04592586308717728, 0.0031665...
441
441
['Paul Honeine']
1411.0161v1
In signal analysis and synthesis, linear approximation theory considers a linear decomposition of any given signal in a set of atoms, collected into a so-called dictionary. Relevant sparse representations are obtained by relaxing the orthogonality condition of the atoms, yielding overcomplete dictionaries with an exten...
Entropy of Overcomplete Kernel Dictionaries
2,014
http://arxiv.org/pdf/1411.0161v1
Title Entropy Overcomplete Kernel Dictionaries Summary signal analysis synthesis linear approximation theory considers linear decomposition given signal set atom collected socalled dictionary Relevant sparse representation obtained relaxing orthogonality condition atom yielding overcomplete dictionary extended number a...
[-0.04676356166601181, 0.012102759443223476, -7.161893881857395e-05, 0.022028762847185135, 0.007572753354907036, -0.018062887713313103, -0.006840858142822981, 0.03293129429221153, -0.07297536730766296, -0.006776285357773304, 0.028882235288619995, 0.03125356510281563, 0.027265353128314018, 0.024965962395071983, -0.01227...
442
442
['Sander Dieleman', 'Kyle W. Willett', 'Joni Dambre']
1503.07077v1
Measuring the morphological parameters of galaxies is a key requirement for studying their formation and evolution. Surveys such as the Sloan Digital Sky Survey (SDSS) have resulted in the availability of very large collections of images, which have permitted population-wide analyses of galaxy morphology. Morphological...
Rotation-invariant convolutional neural networks for galaxy morphology prediction
2,015
http://arxiv.org/pdf/1503.07077v1
Title Rotationinvariant convolutional neural network galaxy morphology prediction Summary Measuring morphological parameter galaxy key requirement studying formation evolution Surveys Sloan Digital Sky Survey SDSS resulted availability large collection image permitted populationwide analysis galaxy morphology Morpholog...
[0.03055027313530445, -0.02055300958454609, -0.008313288912177086, 0.007910335436463356, -0.0001570279273437336, 0.0033753109164536, 0.049835000187158585, 0.009784825146198273, -0.0711127445101738, 0.025293055921792984, -0.02783234603703022, -0.02360847406089306, 0.0166450385004282, -0.011844807304441929, -0.0017254312...
443
443
['Fei Zhu', 'Paul Honeine', 'Maya Kallas']
1407.4420v2
The nonnegative matrix factorization (NMF) is widely used in signal and image processing, including bio-informatics, blind source separation and hyperspectral image analysis in remote sensing. A great challenge arises when dealing with a nonlinear formulation of the NMF. Within the framework of kernel machines, the mod...
Kernel Nonnegative Matrix Factorization Without the Curse of the Pre-image - Application to Unmixing Hyperspectral Images
2,014
http://arxiv.org/pdf/1407.4420v2
Title Kernel Nonnegative Matrix Factorization Without Curse Preimage Application Unmixing Hyperspectral Images Summary nonnegative matrix factorization NMF widely used signal image processing including bioinformatics blind source separation hyperspectral image analysis remote sensing great challenge arises dealing nonl...
[0.0001514033938292414, -0.00814250111579895, -0.006567203905433416, 0.03778566047549248, 0.03802884742617607, -0.014019215479493141, -0.0038603588473051786, 0.017052078619599342, -0.035356305539608, -0.00044711376540362835, -0.006089904811233282, 0.06272603571414948, 0.04076651856303215, 0.058199603110551834, -0.01359...
444
444
['Paul Honeine']
1409.6046v1
Many machine learning frameworks, such as resource-allocating networks, kernel-based methods, Gaussian processes, and radial-basis-function networks, require a sparsification scheme in order to address the online learning paradigm. For this purpose, several online sparsification criteria have been proposed to restrict ...
Approximation errors of online sparsification criteria
2,014
http://arxiv.org/pdf/1409.6046v1
Title Approximation error online sparsification criterion Summary Many machine learning framework resourceallocating network kernelbased method Gaussian process radialbasisfunction network require sparsification scheme order address online learning paradigm purpose several online sparsification criterion proposed restr...
[-0.031394410878419876, 0.04585995152592659, -0.023531474173069, 0.027557507157325745, 0.022309741005301476, -0.027468828484416008, 0.04130548611283302, 0.01737407222390175, -0.037962645292282104, 0.02389407530426979, 0.03831920400261879, 0.019206129014492035, 0.05357345566153526, 0.03625645488500595, -0.02155859395861...
445
445
['Thomas Wiatowski', 'Michael Tschannen', 'Aleksandar Stanić', 'Philipp Grohs', 'Helmut Bölcskei']
1605.08283v1
First steps towards a mathematical theory of deep convolutional neural networks for feature extraction were made---for the continuous-time case---in Mallat, 2012, and Wiatowski and B\"olcskei, 2015. This paper considers the discrete case, introduces new convolutional neural network architectures, and proposes a mathema...
Discrete Deep Feature Extraction: A Theory and New Architectures
2,016
http://arxiv.org/pdf/1605.08283v1
Title Discrete Deep Feature Extraction Theory New Architectures Summary First step towards mathematical theory deep convolutional neural network feature extraction madefor continuoustime casein Mallat 2012 Wiatowski Bolcskei 2015 paper considers discrete case introduces new convolutional neural network architecture pro...
[0.0036944542080163956, 0.035360075533390045, -0.01799187995493412, 0.0591338612139225, 0.007811818737536669, 0.011754624545574188, 0.07544642686843872, 0.013092435896396637, -0.021839654073119164, 0.020087871700525284, 0.03943182900547981, -0.00460813008248806, 0.011793920770287514, 0.0609690397977829, 0.0283645372837...
446
446
['Lifeng Shang', 'Zhengdong Lu', 'Hang Li']
1503.02364v2
We propose Neural Responding Machine (NRM), a neural network-based response generator for Short-Text Conversation. NRM takes the general encoder-decoder framework: it formalizes the generation of response as a decoding process based on the latent representation of the input text, while both encoding and decoding are re...
Neural Responding Machine for Short-Text Conversation
2,015
http://arxiv.org/pdf/1503.02364v2
Title Neural Responding Machine ShortText Conversation Summary propose Neural Responding Machine NRM neural networkbased response generator ShortText Conversation NRM take general encoderdecoder framework formalizes generation response decoding process based latent representation input text encoding decoding realized r...
[0.09230727702379227, 0.020093947649002075, 0.005071334540843964, 0.04573854058980942, -0.011932484805583954, 0.014874564483761787, -0.03324092924594879, -0.00833804626017809, -0.004842789843678474, -0.04736252501606941, 0.022856203839182854, -0.007944663986563683, -0.01691340282559395, 0.062297068536281586, -0.0084837...
447
447
['Nabiha Asghar', 'Pascal Poupart', 'Xin Jiang', 'Hang Li']
1612.03929v5
We propose an online, end-to-end, neural generative conversational model for open-domain dialogue. It is trained using a unique combination of offline two-phase supervised learning and online human-in-the-loop active learning. While most existing research proposes offline supervision or hand-crafted reward functions fo...
Deep Active Learning for Dialogue Generation
2,016
http://arxiv.org/pdf/1612.03929v5
Title Deep Active Learning Dialogue Generation Summary propose online endtoend neural generative conversational model opendomain dialogue trained using unique combination offline twophase supervised learning online humanintheloop active learning existing research proposes offline supervision handcrafted reward function...
[0.0674745962023735, 0.03524432331323624, -0.004978268872946501, -0.04062667116522789, 0.012914099730551243, 0.01766110025346279, 0.02251812256872654, -0.014819574542343616, 0.029238346964120865, -0.049489010125398636, 0.0007450484554283321, -0.025146983563899994, -0.032757505774497986, 0.1011180505156517, 0.0071744625...
448
448
['Karl Moritz Hermann', 'Tomáš Kočiský', 'Edward Grefenstette', 'Lasse Espeholt', 'Will Kay', 'Mustafa Suleyman', 'Phil Blunsom']
1506.03340v3
Teaching machines to read natural language documents remains an elusive challenge. Machine reading systems can be tested on their ability to answer questions posed on the contents of documents that they have seen, but until now large scale training and test datasets have been missing for this type of evaluation. In thi...
Teaching Machines to Read and Comprehend
2,015
http://arxiv.org/pdf/1506.03340v3
Title Teaching Machines Read Comprehend Summary Teaching machine read natural language document remains elusive challenge Machine reading system tested ability answer question posed content document seen large scale training test datasets missing type evaluation work define new methodology resolve bottleneck provides l...
[0.04625609517097473, -0.012580386362969875, -0.0031117768958210945, 0.05506395176053047, -0.008554920554161072, 0.014452523551881313, 0.04525618627667427, -0.025313731282949448, 0.010950825177133083, -0.037284497171640396, -0.014430267736315727, 0.00496969697996974, 0.015908658504486084, 0.042184483259916306, 0.017601...
449
449
['Jianpeng Cheng', 'Dimitri Kartsaklis']
1508.02354v2
Deep compositional models of meaning acting on distributional representations of words in order to produce vectors of larger text constituents are evolving to a popular area of NLP research. We detail a compositional distributional framework based on a rich form of word embeddings that aims at facilitating the interact...
Syntax-Aware Multi-Sense Word Embeddings for Deep Compositional Models of Meaning
2,015
http://arxiv.org/pdf/1508.02354v2
Title SyntaxAware MultiSense Word Embeddings Deep Compositional Models Meaning Summary Deep compositional model meaning acting distributional representation word order produce vector larger text constituent evolving popular area NLP research detail compositional distributional framework based rich form word embeddings ...
[0.02820216491818428, 0.01467214897274971, -0.020512329414486885, 0.049273815006017685, -0.019087502732872963, 0.01134074479341507, -0.03433859720826149, -0.05767248570919037, -0.03139980137348175, -0.05740887299180031, 0.030665528029203415, -0.029023347422480583, 0.048987794667482376, 0.047548871487379074, 0.024283042...
450
450
['Shengxian Wan', 'Yanyan Lan', 'Jiafeng Guo', 'Jun Xu', 'Liang Pang', 'Xueqi Cheng']
1511.08277v1
Matching natural language sentences is central for many applications such as information retrieval and question answering. Existing deep models rely on a single sentence representation or multiple granularity representations for matching. However, such methods cannot well capture the contextualized local information in...
A Deep Architecture for Semantic Matching with Multiple Positional Sentence Representations
2,015
http://arxiv.org/pdf/1511.08277v1
Title Deep Architecture Semantic Matching Multiple Positional Sentence Representations Summary Matching natural language sentence central many application information retrieval question answering Existing deep model rely single sentence representation multiple granularity representation matching However method cannot w...
[0.03854307904839516, 0.0361115038394928, -0.009932505898177624, 0.07326403260231018, -0.04087432473897934, 0.002543353708460927, -0.003332993946969509, -0.0274368766695261, -0.016328807920217514, -0.05835443735122681, -0.00845292117446661, -0.010525465942919254, -0.013230201788246632, 0.016002319753170013, 0.021724427...
451
451
['Yiming Cui', 'Shijin Wang', 'Jianfeng Li']
1512.00177v3
Artificial neural networks are powerful models, which have been widely applied into many aspects of machine translation, such as language modeling and translation modeling. Though notable improvements have been made in these areas, the reordering problem still remains a challenge in statistical machine translations. In...
LSTM Neural Reordering Feature for Statistical Machine Translation
2,015
http://arxiv.org/pdf/1512.00177v3
Title LSTM Neural Reordering Feature Statistical Machine Translation Summary Artificial neural network powerful model widely applied many aspect machine translation language modeling translation modeling Though notable improvement made area reordering problem still remains challenge statistical machine translation pape...
[0.01523587480187416, 0.0033701767679303885, 0.001420486718416214, 0.0695304200053215, -0.0205757487565279, -0.013303345069289207, -0.0031198582146316767, -0.013461161404848099, -0.025361331179738045, -0.00802953913807869, -0.011103509925305843, -0.06025703623890877, 0.08550680428743362, -0.004754986613988876, 0.033382...
452
452
['Shuohang Wang', 'Jing Jiang']
1512.08849v2
Natural language inference (NLI) is a fundamentally important task in natural language processing that has many applications. The recently released Stanford Natural Language Inference (SNLI) corpus has made it possible to develop and evaluate learning-centered methods such as deep neural networks for natural language i...
Learning Natural Language Inference with LSTM
2,015
http://arxiv.org/pdf/1512.08849v2
Title Learning Natural Language Inference LSTM Summary Natural language inference NLI fundamentally important task natural language processing many application recently released Stanford Natural Language Inference SNLI corpus made possible develop evaluate learningcentered method deep neural network natural language in...
[0.05412283167243004, 0.020508384332060814, -0.0009332384215667844, 0.0482247993350029, -0.05180315300822258, 0.00462347362190485, 0.027472184970974922, -0.001867720391601324, 0.036760490387678146, -0.050992630422115326, 0.012076687067747116, 0.0026270465459674597, 0.014237107709050179, 0.03020339272916317, 0.010326889...
453
453
['Phong Le', 'Willem Zuidema']
1603.00423v1
Recursive neural networks (RNN) and their recently proposed extension recursive long short term memory networks (RLSTM) are models that compute representations for sentences, by recursively combining word embeddings according to an externally provided parse tree. Both models thus, unlike recurrent networks, explicitly ...
Quantifying the vanishing gradient and long distance dependency problem in recursive neural networks and recursive LSTMs
2,016
http://arxiv.org/pdf/1603.00423v1
Title Quantifying vanishing gradient long distance dependency problem recursive neural network recursive LSTMs Summary Recursive neural network RNN recently proposed extension recursive long short term memory network RLSTM model compute representation sentence recursively combining word embeddings according externally ...
[0.03300371766090393, 0.015306524001061916, 0.01262767892330885, 0.06967125833034515, -0.03792081028223038, -0.01024245098233223, -0.047460637986660004, 0.026505716145038605, -0.008483223617076874, -0.0677141547203064, -0.003822858678176999, -0.014204154722392559, 0.033913105726242065, -0.007055546622723341, -0.0216247...
454
454
['Yang Liu', 'Sujian Li', 'Xiaodong Zhang', 'Zhifang Sui']
1603.02776v1
Without discourse connectives, classifying implicit discourse relations is a challenging task and a bottleneck for building a practical discourse parser. Previous research usually makes use of one kind of discourse framework such as PDTB or RST to improve the classification performance on discourse relations. Actually,...
Implicit Discourse Relation Classification via Multi-Task Neural Networks
2,016
http://arxiv.org/pdf/1603.02776v1
Title Implicit Discourse Relation Classification via MultiTask Neural Networks Summary Without discourse connective classifying implicit discourse relation challenging task bottleneck building practical discourse parser Previous research usually make use one kind discourse framework PDTB RST improve classification perf...
[0.03985503688454628, 0.014695283025503159, 0.005900284741073847, 0.05481695383787155, -0.01843344420194626, 0.01809452660381794, 0.027498049661517143, -0.02179475873708725, 0.004185981582850218, -0.09269911050796509, -0.0570853017270565, 0.01660793460905552, -0.0031876862049102783, -0.0018962051253765821, -0.025644404...
455
455
['Peng Li', 'Heng Huang']
1603.09405v1
Neural network based approaches for sentence relation modeling automatically generate hidden matching features from raw sentence pairs. However, the quality of matching feature representation may not be satisfied due to complex semantic relations such as entailment or contradiction. To address this challenge, we propos...
Enhancing Sentence Relation Modeling with Auxiliary Character-level Embedding
2,016
http://arxiv.org/pdf/1603.09405v1
Title Enhancing Sentence Relation Modeling Auxiliary Characterlevel Embedding Summary Neural network based approach sentence relation modeling automatically generate hidden matching feature raw sentence pair However quality matching feature representation may satisfied due complex semantic relation entailment contradic...
[0.045276906341314316, 0.07869525253772736, -0.01902376301586628, 0.08849707990884781, -0.04558378458023071, -0.014144666492938995, -0.01226112898439169, 0.008166917599737644, 0.027594882994890213, -0.029279377311468124, 0.002444598823785782, -0.0108577786013484, -0.008450926281511784, 0.02788546308875084, -0.002628518...
456
456
['Byungsoo Kim', 'Hwanjo Yu', 'Gary Geunbae Lee']
1605.07918v1
Previous studies in Open Information Extraction (Open IE) are mainly based on extraction patterns. They manually define patterns or automatically learn them from a large corpus. However, these approaches are limited when grasping the context of a sentence, and they fail to capture implicit relations. In this paper, we ...
Automatic Open Knowledge Acquisition via Long Short-Term Memory Networks with Feedback Negative Sampling
2,016
http://arxiv.org/pdf/1605.07918v1
Title Automatic Open Knowledge Acquisition via Long ShortTerm Memory Networks Feedback Negative Sampling Summary Previous study Open Information Extraction Open IE mainly based extraction pattern manually define pattern automatically learn large corpus However approach limited grasping context sentence fail capture imp...
[0.032870106399059296, 0.010768353007733822, 0.016338851302862167, 0.07687193155288696, -0.04875437170267105, -0.01448119431734085, -0.021911071613430977, 0.034561749547719955, -0.0004447177634574473, -0.07236410677433014, -0.04646538197994232, -0.0294781643897295, -0.0017716458532959223, 0.05546487495303154, -0.036198...
457
457
['Yuanzhe Zhang', 'Kang Liu', 'Shizhu He', 'Guoliang Ji', 'Zhanyi Liu', 'Hua Wu', 'Jun Zhao']
1606.00979v1
With the rapid growth of knowledge bases (KBs) on the web, how to take full advantage of them becomes increasingly important. Knowledge base-based question answering (KB-QA) is one of the most promising approaches to access the substantial knowledge. Meantime, as the neural network-based (NN-based) methods develop, NN-...
Question Answering over Knowledge Base with Neural Attention Combining Global Knowledge Information
2,016
http://arxiv.org/pdf/1606.00979v1
Title Question Answering Knowledge Base Neural Attention Combining Global Knowledge Information Summary rapid growth knowledge base KBs web take full advantage becomes increasingly important Knowledge basebased question answering KBQA one promising approach access substantial knowledge Meantime neural networkbased NNba...
[0.06426072120666504, 0.007134913932532072, -0.027330251410603523, 0.04187152162194252, 0.0013349675573408604, -0.00547182047739625, -0.014516432769596577, 0.013532851822674274, 0.0057568978518247604, -0.05016082152724266, 0.013598648831248283, -0.011826754547655582, -0.04393204674124718, 0.03993247076869011, 0.0685734...
458
458
['Vladyslav Kolesnyk', 'Tim Rocktäschel', 'Sebastian Riedel']
1606.01404v1
The ability to reason with natural language is a fundamental prerequisite for many NLP tasks such as information extraction, machine translation and question answering. To quantify this ability, systems are commonly tested whether they can recognize textual entailment, i.e., whether one sentence can be inferred from an...
Generating Natural Language Inference Chains
2,016
http://arxiv.org/pdf/1606.01404v1
Title Generating Natural Language Inference Chains Summary ability reason natural language fundamental prerequisite many NLP task information extraction machine translation question answering quantify ability system commonly tested whether recognize textual entailment ie whether one sentence inferred another one Howeve...
[0.06048582121729851, 0.025336531922221184, -0.0040178438648581505, 0.04967367276549339, -0.08092920482158661, 0.008150935173034668, -0.006997069343924522, -0.012140650302171707, 0.04903362691402435, -0.028601769357919693, 0.03629592806100845, 0.10279527306556702, -0.006563949398696423, 0.031010763719677925, 0.02452779...
459
459
['Dirk Weissenborn', 'Tim Rocktäschel']
1606.03002v1
Recurrent neural networks such as the GRU and LSTM found wide adoption in natural language processing and achieve state-of-the-art results for many tasks. These models are characterized by a memory state that can be written to and read from by applying gated composition operations to the current input and the previous ...
MuFuRU: The Multi-Function Recurrent Unit
2,016
http://arxiv.org/pdf/1606.03002v1
Title MuFuRU MultiFunction Recurrent Unit Summary Recurrent neural network GRU LSTM found wide adoption natural language processing achieve stateoftheart result many task model characterized memory state written read applying gated composition operation current input previous state However cover small subset potentiall...
[0.03487687185406685, 0.019420964643359184, 0.0063804457895457745, 0.012132386676967144, -0.05353996157646179, 0.0033452834468334913, 0.018502071499824524, -0.0281940009444952, 0.008634557016193867, -0.06641259044408798, 0.04120872542262077, -0.06622032076120377, 0.04509202390909195, 0.0946258157491684, 0.0180994383990...
460
460
['Hendrik Strobelt', 'Sebastian Gehrmann', 'Hanspeter Pfister', 'Alexander M. Rush']
1606.07461v2
Recurrent neural networks, and in particular long short-term memory (LSTM) networks, are a remarkably effective tool for sequence modeling that learn a dense black-box hidden representation of their sequential input. Researchers interested in better understanding these models have studied the changes in hidden state re...
LSTMVis: A Tool for Visual Analysis of Hidden State Dynamics in Recurrent Neural Networks
2,016
http://arxiv.org/pdf/1606.07461v2
Title LSTMVis Tool Visual Analysis Hidden State Dynamics Recurrent Neural Networks Summary Recurrent neural network particular long shortterm memory LSTM network remarkably effective tool sequence modeling learn dense blackbox hidden representation sequential input Researchers interested better understanding model stud...
[-0.006692376919090748, -0.013991089537739754, -0.043199650943279266, 0.05151674896478653, -0.03188309445977211, -0.030941110104322433, 0.008468410931527615, 0.04561137780547142, -0.06617894768714905, 0.004464995581656694, -0.0233839713037014, 0.0016762190498411655, 0.04935676231980324, 0.045944906771183014, 0.01374636...
461
461
['Abigail See', 'Minh-Thang Luong', 'Christopher D. Manning']
1606.09274v1
Neural Machine Translation (NMT), like many other deep learning domains, typically suffers from over-parameterization, resulting in large storage sizes. This paper examines three simple magnitude-based pruning schemes to compress NMT models, namely class-blind, class-uniform, and class-distribution, which differ in ter...
Compression of Neural Machine Translation Models via Pruning
2,016
http://arxiv.org/pdf/1606.09274v1
Title Compression Neural Machine Translation Models via Pruning Summary Neural Machine Translation NMT like many deep learning domain typically suffers overparameterization resulting large storage size paper examines three simple magnitudebased pruning scheme compress NMT model namely classblind classuniform classdistr...
[-0.013870018534362316, 0.03950463607907295, -0.0017748986138030887, 0.03428439423441887, -0.017992300912737846, 0.00034034220152534544, 0.008371628820896149, 0.044367920607328415, -0.0744989737868309, 0.02078443393111229, -0.024357156828045845, -0.02196624130010605, 0.0016537809278815985, -0.02828974835574627, 0.01635...
462
462
['Janez Starc', 'Dunja Mladenić']
1607.06025v2
Natural Language Inference is an important task for Natural Language Understanding. It is concerned with classifying the logical relation between two sentences. In this paper, we propose several text generative neural networks for generating text hypothesis, which allows construction of new Natural Language Inference d...
Constructing a Natural Language Inference Dataset using Generative Neural Networks
2,016
http://arxiv.org/pdf/1607.06025v2
Title Constructing Natural Language Inference Dataset using Generative Neural Networks Summary Natural Language Inference important task Natural Language Understanding concerned classifying logical relation two sentence paper propose several text generative neural network generating text hypothesis allows construction ...
[0.05523627623915672, 0.07741091400384903, 0.0064531653188169, 0.047283172607421875, -0.06524530053138733, 0.014730148017406464, 0.01783418469130993, -0.008247275836765766, 0.04907894507050514, -0.01197761856019497, -0.003073606174439192, 0.015673432499170303, 0.014606429263949394, 0.027546612545847893, 0.0313973799347...
463
463
['Peng Li', 'Wei Li', 'Zhengyan He', 'Xuguang Wang', 'Ying Cao', 'Jie Zhou', 'Wei Xu']
1607.06275v2
While question answering (QA) with neural network, i.e. neural QA, has achieved promising results in recent years, lacking of large scale real-word QA dataset is still a challenge for developing and evaluating neural QA system. To alleviate this problem, we propose a large scale human annotated real-world QA dataset We...
Dataset and Neural Recurrent Sequence Labeling Model for Open-Domain Factoid Question Answering
2,016
http://arxiv.org/pdf/1607.06275v2
Title Dataset Neural Recurrent Sequence Labeling Model OpenDomain Factoid Question Answering Summary question answering QA neural network ie neural QA achieved promising result recent year lacking large scale realword QA dataset still challenge developing evaluating neural QA system alleviate problem propose large scal...
[0.0691339522600174, 0.0513947531580925, 0.0172710157930851, 0.021274419501423836, 0.009584801271557808, 0.012726183980703354, 0.02850259840488434, -0.006918837316334248, 0.015944670885801315, -0.03456156328320503, 0.009860329329967499, -0.01773625984787941, -0.016980916261672974, 0.0616447851061821, 0.0348480343818664...
464
464
['Soroush Vosoughi', 'Prashanth Vijayaraghavan', 'Deb Roy']
1607.07514v1
We present Tweet2Vec, a novel method for generating general-purpose vector representation of tweets. The model learns tweet embeddings using character-level CNN-LSTM encoder-decoder. We trained our model on 3 million, randomly selected English-language tweets. The model was evaluated using two methods: tweet semantic s...
Tweet2Vec: Learning Tweet Embeddings Using Character-level CNN-LSTM Encoder-Decoder
2,016
http://arxiv.org/pdf/1607.07514v1
Title Tweet2Vec Learning Tweet Embeddings Using Characterlevel CNNLSTM EncoderDecoder Summary present Tweet2Vec novel method generating generalpurpose vector representation tweet model learns tweet embeddings using characterlevel CNNLSTM encoderdecoder trained model 3 million randomly selected Englishlanguage tweet mod...
[0.04644796624779701, 0.02613394893705845, 0.01620151475071907, 0.04045052453875542, -0.0040725297294557095, 0.06053799018263817, -0.009877495467662811, 0.023585766553878784, 0.0002846387214958668, -0.023674432188272476, 0.032815784215927124, -0.017511209473013878, -0.05169404298067093, 0.03193754330277443, 0.030918573...
465
465
['Lei Yu', 'Jan Buys', 'Phil Blunsom']
1609.08194v1
We introduce an online neural sequence to sequence model that learns to alternate between encoding and decoding segments of the input as it is read. By independently tracking the encoding and decoding representations our algorithm permits exact polynomial marginalization of the latent segmentation during training, and ...
Online Segment to Segment Neural Transduction
2,016
http://arxiv.org/pdf/1609.08194v1
Title Online Segment Segment Neural Transduction Summary introduce online neural sequence sequence model learns alternate encoding decoding segment input read independently tracking encoding decoding representation algorithm permit exact polynomial marginalization latent segmentation training decoding beam search emplo...
[0.014687427319586277, 0.08479885011911392, 0.022117841988801956, 0.019355658441781998, -0.021395109593868256, -0.004048784729093313, -0.001354938605800271, -0.010278400033712387, -0.08875929564237595, -0.012650947086513042, -0.002195000648498535, -0.02880951762199402, 0.031131576746702194, 0.09306752681732178, 0.03184...
466
466
['Tomáš Kočiský', 'Gábor Melis', 'Edward Grefenstette', 'Chris Dyer', 'Wang Ling', 'Phil Blunsom', 'Karl Moritz Hermann']
1609.09315v1
We present a novel semi-supervised approach for sequence transduction and apply it to semantic parsing. The unsupervised component is based on a generative model in which latent sentences generate the unpaired logical forms. We apply this method to a number of semantic parsing tasks focusing on domains with limited acc...
Semantic Parsing with Semi-Supervised Sequential Autoencoders
2,016
http://arxiv.org/pdf/1609.09315v1
Title Semantic Parsing SemiSupervised Sequential Autoencoders Summary present novel semisupervised approach sequence transduction apply semantic parsing unsupervised component based generative model latent sentence generate unpaired logical form apply method number semantic parsing task focusing domain limited access l...
[0.018922995775938034, 0.0721251368522644, -0.022611845284700394, 0.04588526487350464, -0.03592238202691078, 0.013234042562544346, 0.02491137385368347, -0.060034338384866714, -0.0037507028318941593, -0.05410213768482208, 0.02314099483191967, 0.021151628345251083, -0.01956808753311634, 0.09992970526218414, 0.00460032233...
467
467
['Lina M. Rojas Barahona', 'Milica Gasic', 'Nikola Mrkšić', 'Pei-Hao Su', 'Stefan Ultes', 'Tsung-Hsien Wen', 'Steve Young']
1610.04120v1
This paper presents a deep learning architecture for the semantic decoder component of a Statistical Spoken Dialogue System. In a slot-filling dialogue, the semantic decoder predicts the dialogue act and a set of slot-value pairs from a set of n-best hypotheses returned by the Automatic Speech Recognition. Most current...
Exploiting Sentence and Context Representations in Deep Neural Models for Spoken Language Understanding
2,016
http://arxiv.org/pdf/1610.04120v1
Title Exploiting Sentence Context Representations Deep Neural Models Spoken Language Understanding Summary paper present deep learning architecture semantic decoder component Statistical Spoken Dialogue System slotfilling dialogue semantic decoder predicts dialogue act set slotvalue pair set nbest hypothesis returned A...
[0.025100350379943848, 0.030358703806996346, 0.002529171761125326, 0.06476062536239624, -0.023387886583805084, 0.012930279597640038, 0.028537053614854813, -0.02853717841207981, -0.04127072915434837, -0.09590990096330643, -0.038283973932266235, -0.009800244122743607, 0.015018795616924763, 0.04422890394926071, -0.0282749...
468
468
['Lei Yu', 'Phil Blunsom', 'Chris Dyer', 'Edward Grefenstette', 'Tomas Kocisky']
1611.02554v2
We formulate sequence to sequence transduction as a noisy channel decoding problem and use recurrent neural networks to parameterise the source and channel models. Unlike direct models which can suffer from explaining-away effects during training, noisy channel models must produce outputs that explain their inputs, and...
The Neural Noisy Channel
2,016
http://arxiv.org/pdf/1611.02554v2
Title Neural Noisy Channel Summary formulate sequence sequence transduction noisy channel decoding problem use recurrent neural network parameterise source channel model Unlike direct model suffer explainingaway effect training noisy channel model must produce output explain input component model trained paired trainin...
[0.0011456308420747519, 0.06471269577741623, 0.021822137758135796, 0.01673102006316185, 0.013105972670018673, 0.012457650154829025, -0.027519145980477333, 0.00708532752469182, -0.12257658690214157, 0.004137362819164991, -0.014020216651260853, -4.6241359086707234e-05, 0.03299223631620407, 0.03300931304693222, 0.02741978...
469
469
['Iulian Vlad Serban', 'Ryan Lowe', 'Laurent Charlin', 'Joelle Pineau']
1611.06216v1
Researchers have recently started investigating deep neural networks for dialogue applications. In particular, generative sequence-to-sequence (Seq2Seq) models have shown promising results for unstructured tasks, such as word-level dialogue response generation. The hope is that such models will be able to leverage mass...
Generative Deep Neural Networks for Dialogue: A Short Review
2,016
http://arxiv.org/pdf/1611.06216v1
Title Generative Deep Neural Networks Dialogue Short Review Summary Researchers recently started investigating deep neural network dialogue application particular generative sequencetosequence Seq2Seq model shown promising result unstructured task wordlevel dialogue response generation hope model able leverage massive ...
[0.06390269100666046, 0.0733368992805481, -0.011288669891655445, 0.029294809326529503, 0.005731811746954918, -0.008158352226018906, 0.013075281865894794, -0.015365160070359707, -0.004767810460180044, -0.02165645733475685, 0.020351719111204147, -0.03490282595157623, 0.005236675031483173, 0.08828261494636536, 0.004751257...
470
470
['Avishkar Bhoopchand', 'Tim Rocktäschel', 'Earl Barr', 'Sebastian Riedel']
1611.08307v1
To enhance developer productivity, all modern integrated development environments (IDEs) include code suggestion functionality that proposes likely next tokens at the cursor. While current IDEs work well for statically-typed languages, their reliance on type annotations means that they do not provide the same level of ...
Learning Python Code Suggestion with a Sparse Pointer Network
2,016
http://arxiv.org/pdf/1611.08307v1
Title Learning Python Code Suggestion Sparse Pointer Network Summary enhance developer productivity modern integrated development environment IDEs include code suggestion functionality proposes likely next token cursor current IDEs work well staticallytyped language reliance type annotation mean provide level support d...
[0.04339522495865822, 0.024847663938999176, -0.014519850723445415, 0.019994234666228294, -0.00523100420832634, -0.015087258070707321, -0.006111054215580225, 0.03958819806575775, 0.0005323246587067842, -0.01784777268767357, -0.015427467413246632, 0.033090364187955856, 0.0035767473746091127, 0.05823474004864693, 0.026940...
471
471
['Guillaume Klein', 'Yoon Kim', 'Yuntian Deng', 'Jean Senellart', 'Alexander M. Rush']
1701.02810v2
We describe an open-source toolkit for neural machine translation (NMT). The toolkit prioritizes efficiency, modularity, and extensibility with the goal of supporting NMT research into model architectures, feature representations, and source modalities, while maintaining competitive performance and reasonable training ...
OpenNMT: Open-Source Toolkit for Neural Machine Translation
2,017
http://arxiv.org/pdf/1701.02810v2
Title OpenNMT OpenSource Toolkit Neural Machine Translation Summary describe opensource toolkit neural machine translation NMT toolkit prioritizes efficiency modularity extensibility goal supporting NMT research model architecture feature representation source modality maintaining competitive performance reasonable tra...
[0.012157713994383812, 0.026993492618203163, -0.00046563352225348353, 0.02612263523042202, -0.035118117928504944, 0.012179067358374596, 0.02470211684703827, 0.02371455542743206, -0.05222651734948158, -0.014534087851643562, -0.015913058072328568, -0.021140730008482933, 0.040106549859046936, 0.0255400612950325, 0.0481840...
472
472
['Dirk Weissenborn', 'Georg Wiese', 'Laura Seiffe']
1703.04816v3
Recent development of large-scale question answering (QA) datasets triggered a substantial amount of research into end-to-end neural architectures for QA. Increasingly complex systems have been conceived without comparison to simpler neural baseline systems that would justify their complexity. In this work, we propose ...
Making Neural QA as Simple as Possible but not Simpler
2,017
http://arxiv.org/pdf/1703.04816v3
Title Making Neural QA Simple Possible Simpler Summary Recent development largescale question answering QA datasets triggered substantial amount research endtoend neural architecture QA Increasingly complex system conceived without comparison simpler neural baseline system would justify complexity work propose simple h...
[0.06591969728469849, 0.029870105907320976, -0.015024044550955296, 0.04606082662940025, 0.0002484614960849285, 0.008769730105996132, -0.015431920997798443, -0.008799200877547264, -0.0110649224370718, -0.033688630908727646, -0.002634844509884715, -0.02954694628715515, -0.02410193532705307, 0.06254766881465912, 0.0399308...
473
473
['Albert Gatt', 'Emiel Krahmer']
1703.09902v4
This paper surveys the current state of the art in Natural Language Generation (NLG), defined as the task of generating text or speech from non-linguistic input. A survey of NLG is timely in view of the changes that the field has undergone over the past decade or so, especially in relation to new (usually data-driven) ...
Survey of the State of the Art in Natural Language Generation: Core tasks, applications and evaluation
2,017
http://arxiv.org/pdf/1703.09902v4
Title Survey State Art Natural Language Generation Core task application evaluation Summary paper survey current state art Natural Language Generation NLG defined task generating text speech nonlinguistic input survey NLG timely view change field undergone past decade especially relation new usually datadriven method w...
[0.06841406226158142, 0.0017079530516639352, 0.0012855371460318565, 0.028355872258543968, -0.041360627859830856, -0.03688322380185127, 0.03379391133785248, -0.003396128537133336, 0.00891396775841713, -0.057099319994449615, 0.025376342236995697, -0.010400498285889626, 0.01604187861084938, 0.04415862634778023, 0.00330479...
474
474
['Yaoyuan Zhang', 'Zhenxu Ye', 'Yansong Feng', 'Dongyan Zhao', 'Rui Yan']
1704.02312v1
Sentence simplification reduces semantic complexity to benefit people with language impairments. Previous simplification studies on the sentence level and word level have achieved promising results but also meet great challenges. For sentence-level studies, sentences after simplification are fluent but sometimes are no...
A Constrained Sequence-to-Sequence Neural Model for Sentence Simplification
2,017
http://arxiv.org/pdf/1704.02312v1
Title Constrained SequencetoSequence Neural Model Sentence Simplification Summary Sentence simplification reduces semantic complexity benefit people language impairment Previous simplification study sentence level word level achieved promising result also meet great challenge sentencelevel study sentence simplification...
[0.037985920906066895, 0.08970826864242554, -0.02927764318883419, 0.006693972274661064, -0.04820210114121437, -0.029439451172947884, 0.030323663726449013, 0.007917067036032677, -0.03963814303278923, -0.009375032037496567, 0.036269597709178925, 0.005774978548288345, 0.03666891157627106, 0.013577942736446857, 0.043408662...
475
475
['Mo Yu', 'Wenpeng Yin', 'Kazi Saidul Hasan', 'Cicero dos Santos', 'Bing Xiang', 'Bowen Zhou']
1704.06194v2
Relation detection is a core component for many NLP applications including Knowledge Base Question Answering (KBQA). In this paper, we propose a hierarchical recurrent neural network enhanced by residual learning that detects KB relations given an input question. Our method uses deep residual bidirectional LSTMs to com...
Improved Neural Relation Detection for Knowledge Base Question Answering
2,017
http://arxiv.org/pdf/1704.06194v2
Title Improved Neural Relation Detection Knowledge Base Question Answering Summary Relation detection core component many NLP application including Knowledge Base Question Answering KBQA paper propose hierarchical recurrent neural network enhanced residual learning detects KB relation given input question method us dee...
[0.06825186312198639, 0.008547118864953518, 0.010991828516125679, 0.10218099504709244, -0.004965547937899828, 0.02915715053677559, -0.0367828905582428, 0.04724039137363434, 0.02808644063770771, -0.043460968881845474, -0.006258853245526552, -0.00458939652889967, -0.03285800665616989, 0.0028230154421180487, 0.00364264682...
476
476
['Edwin Simonnet', 'Sahar Ghannay', 'Nathalie Camelin', 'Yannick Estève', 'Renato De Mori']
1705.09515v1
This paper addresses the problem of automatic speech recognition (ASR) error detection and their use for improving spoken language understanding (SLU) systems. In this study, the SLU task consists in automatically extracting, from ASR transcriptions , semantic concepts and concept/values pairs in a e.g touristic inform...
ASR error management for improving spoken language understanding
2,017
http://arxiv.org/pdf/1705.09515v1
Title ASR error management improving spoken language understanding Summary paper address problem automatic speech recognition ASR error detection use improving spoken language understanding SLU system study SLU task consists automatically extracting ASR transcription semantic concept conceptvalues pair eg touristic inf...
[0.03429742529988289, 0.023915894329547882, 0.03340352326631546, 0.06692226231098175, 0.042175084352493286, 0.008284401148557663, -0.022708090022206306, 0.011733021587133408, -0.04916686564683914, -0.07305515557527542, -0.026697484776377678, -0.026233786717057228, 0.041347064077854156, 0.022614168003201485, 0.002321220...
477
477
['Dirk Weissenborn', 'Tomáš Kočiský', 'Chris Dyer']
1706.02596v2
Common-sense or background knowledge is required to understand natural language, but in most neural natural language understanding (NLU) systems, the requisite background knowledge is indirectly acquired from static corpora. We develop a new reading architecture for the dynamic integration of explicit background knowle...
Dynamic Integration of Background Knowledge in Neural NLU Systems
2,017
http://arxiv.org/pdf/1706.02596v2
Title Dynamic Integration Background Knowledge Neural NLU Systems Summary Commonsense background knowledge required understand natural language neural natural language understanding NLU system requisite background knowledge indirectly acquired static corpus develop new reading architecture dynamic integration explicit ...
[0.0630718320608139, -0.016112931072711945, -0.01979481801390648, 0.012163262814283371, -0.04035535082221031, 0.03136575594544411, 0.002129240892827511, -0.05056740343570709, 0.04119458422064781, -0.09115668386220932, 0.030752604827284813, 0.025859611108899117, 0.005785737186670303, 0.09818432480096817, 0.0120423212647...
478
478
['Shuai Tang', 'Hailin Jin', 'Chen Fang', 'Zhaowen Wang', 'Virginia R. de Sa']
1706.03146v1
We study the skip-thought model with neighborhood information as weak supervision. More specifically, we propose a skip-thought neighbor model to consider the adjacent sentences as a neighborhood. We train our skip-thought neighbor model on a large corpus with continuous sentences, and then evaluate the trained model o...
Rethinking Skip-thought: A Neighborhood based Approach
2,017
http://arxiv.org/pdf/1706.03146v1
Title Rethinking Skipthought Neighborhood based Approach Summary study skipthought model neighborhood information weak supervision specifically propose skipthought neighbor model consider adjacent sentence neighborhood train skipthought neighbor model large corpus continuous sentence evaluate trained model 7 task inclu...
[0.07042957842350006, 0.026709692552685738, -0.023683154955506325, 0.05327232927083969, -0.08143114298582077, 0.015316244214773178, 0.005118263885378838, 0.02174159325659275, -0.007665194105356932, -0.0659298449754715, -0.010834223590791225, 0.05457272380590439, -0.03386347368359566, -0.02613106369972229, -0.0139488968...
479
479
['Georg Wiese', 'Dirk Weissenborn', 'Mariana Neves']
1706.03610v2
Factoid question answering (QA) has recently benefited from the development of deep learning (DL) systems. Neural network models outperform traditional approaches in domains where large datasets exist, such as SQuAD (ca. 100,000 questions) for Wikipedia articles. However, these systems have not yet been applied to QA i...
Neural Domain Adaptation for Biomedical Question Answering
2,017
http://arxiv.org/pdf/1706.03610v2
Title Neural Domain Adaptation Biomedical Question Answering Summary Factoid question answering QA recently benefited development deep learning DL system Neural network model outperform traditional approach domain large datasets exist SQuAD ca 100000 question Wikipedia article However system yet applied QA specific dom...
[0.04262051731348038, 0.05837785452604294, -0.011497436091303825, -0.008565559983253479, 0.009716584347188473, 0.011876585893332958, 0.03312915936112404, 0.03677397593855858, -0.02092050202190876, -0.03381045162677765, 0.014537196606397629, 0.011613793671131134, -0.026118028908967972, 0.0651947557926178, 0.024646587669...
480
480
['Sandeep Subramanian', 'Tong Wang', 'Xingdi Yuan', 'Saizheng Zhang', 'Adam Trischler', 'Yoshua Bengio']
1706.04560v2
We propose a two-stage neural model to tackle question generation from documents. Our model first estimates the probability that word sequences in a document compose "interesting" answers using a neural model trained on a question-answering corpus. We thus take a data-driven approach to interestingness. Predicted key p...
Neural Models for Key Phrase Detection and Question Generation
2,017
http://arxiv.org/pdf/1706.04560v2
Title Neural Models Key Phrase Detection Question Generation Summary propose twostage neural model tackle question generation document model first estimate probability word sequence document compose interesting answer using neural model trained questionanswering corpus thus take datadriven approach interestingness Pred...
[0.08843544125556946, -0.004547679331153631, 0.000339330465067178, 0.03676971048116684, -0.02507578767836094, 0.02954575978219509, -0.02305273525416851, 0.019697800278663635, 0.019017862156033516, -0.03619661554694176, 0.01438960526138544, -0.011492978781461716, -0.049285974353551865, 0.07849238067865372, 0.03141279518...
481
481
['Georg Wiese', 'Dirk Weissenborn', 'Mariana Neves']
1706.08568v1
This paper describes our submission to the 2017 BioASQ challenge. We participated in Task B, Phase B which is concerned with biomedical question answering (QA). We focus on factoid and list question, using an extractive QA model, that is, we restrict our system to output substrings of the provided text snippets. At the...
Neural Question Answering at BioASQ 5B
2,017
http://arxiv.org/pdf/1706.08568v1
Title Neural Question Answering BioASQ 5B Summary paper describes submission 2017 BioASQ challenge participated Task B Phase B concerned biomedical question answering QA focus factoid list question using extractive QA model restrict system output substring provided text snippet core system use FastQA stateoftheart neur...
[0.06912880390882492, 0.03574235364794731, 0.018872641026973724, 0.007223246153444052, -0.0027934906538575888, 0.009961670264601707, 0.0015119727468118072, 0.036029551178216934, 0.0037093721330165863, -0.004548780154436827, -0.003829070134088397, -0.011537163518369198, -0.01790100336074829, 0.030550070106983185, 0.0534...
482
482
['Hongyu Guo']
1707.01555v1
While natural languages are compositional, how state-of-the-art neural models achieve compositionality is still unclear. We propose a deep network, which not only achieves competitive accuracy for text classification, but also exhibits compositional behavior. That is, while creating hierarchical representations of a pi...
A Deep Network with Visual Text Composition Behavior
2,017
http://arxiv.org/pdf/1707.01555v1
Title Deep Network Visual Text Composition Behavior Summary natural language compositional stateoftheart neural model achieve compositionality still unclear propose deep network achieves competitive accuracy text classification also exhibit compositional behavior creating hierarchical representation piece text sentence...
[0.03845909610390663, 0.029409339651465416, -0.015025964938104153, 0.031386107206344604, -0.046831391751766205, 0.0086111044511199, 0.004960916005074978, 0.0012206637766212225, -0.04155999794602394, -0.017742177471518517, 0.003201203653588891, -0.0288506131619215, 0.044291768223047256, 0.058290015906095505, 0.032687511...
483
483
['Mario Giulianelli']
1708.03910v1
There exist two main approaches to automatically extract affective orientation: lexicon-based and corpus-based. In this work, we argue that these two methods are compatible and show that combining them can improve the accuracy of emotion classifiers. In particular, we introduce a novel variant of the Label Propagation ...
Semi-supervised emotion lexicon expansion with label propagation and specialized word embeddings
2,017
http://arxiv.org/pdf/1708.03910v1
Title Semisupervised emotion lexicon expansion label propagation specialized word embeddings Summary exist two main approach automatically extract affective orientation lexiconbased corpusbased work argue two method compatible show combining improve accuracy emotion classifier particular introduce novel variant Label P...
[0.02880338951945305, 0.05068530887365341, 0.008375333622097969, 0.07251240313053131, -0.009983974508941174, 0.02534208446741104, -0.035031095147132874, -0.03878992050886154, 0.03784232586622238, -0.05756577104330063, -0.045940157026052475, 0.020680682733654976, -0.05398101359605789, 0.009172444231808186, -0.0398169606...
484
484
['Elahe Rahimtoroghi', 'Jiaqi Wu', 'Ruimin Wang', 'Pranav Anand', 'Marilyn A Walker']
1708.09040v1
Many genres of natural language text are narratively structured, a testament to our predilection for organizing our experiences as narratives. There is broad consensus that understanding a narrative requires identifying and tracking the goals and desires of the characters and their narrative outcomes. However, to date,...
Modelling Protagonist Goals and Desires in First-Person Narrative
2,017
http://arxiv.org/pdf/1708.09040v1
Title Modelling Protagonist Goals Desires FirstPerson Narrative Summary Many genre natural language text narratively structured testament predilection organizing experience narrative broad consensus understanding narrative requires identifying tracking goal desire character narrative outcome However date limited work c...
[0.057369738817214966, 0.05930757895112038, -0.02542848140001297, 0.02270907163619995, -0.05294479429721832, 0.004344637040048838, 0.005316389724612236, -0.02398129366338253, 0.0035211159847676754, -0.0430864617228508, 0.053379856050014496, -0.04525022953748703, 0.035953864455223083, 0.06503397971391678, -0.04321665316...
485
485
['Felix Hill', 'Karl Moritz Hermann', 'Phil Blunsom', 'Stephen Clark']
1710.09867v1
Neural network-based systems can now learn to locate the referents of words and phrases in images, answer questions about visual scenes, and even execute symbolic instructions as first-person actors in partially-observable worlds. To achieve this so-called grounded language learning, models must overcome certain well-s...
Understanding Grounded Language Learning Agents
2,017
http://arxiv.org/pdf/1710.09867v1
Title Understanding Grounded Language Learning Agents Summary Neural networkbased system learn locate referent word phrase image answer question visual scene even execute symbolic instruction firstperson actor partiallyobservable world achieve socalled grounded language learning model must overcome certain wellstudied ...
[0.039170317351818085, -0.01116965338587761, -0.02970067784190178, 0.024066317826509476, -0.014153378084301949, 0.00025646082940511405, -0.009390239603817463, -0.029126809909939766, -0.04228420928120613, -0.058828309178352356, -0.010626576840877533, 0.03133980184793472, 0.023358115926384926, 0.06796994805335999, 0.0125...
486
486
['Anjishnu Kumar', 'Arpit Gupta', 'Julian Chan', 'Sam Tucker', 'Bjorn Hoffmeister', 'Markus Dreyer', 'Stanislav Peshterliev', 'Ankur Gandhe', 'Denis Filiminov', 'Ariya Rastrow', 'Christian Monson', 'Agnika Kumar']
1711.00549v4
This paper presents the design of the machine learning architecture that underlies the Alexa Skills Kit (ASK) a large scale Spoken Language Understanding (SLU) Software Development Kit (SDK) that enables developers to extend the capabilities of Amazon's virtual assistant, Alexa. At Amazon, the infrastructure powers ove...
Just ASK: Building an Architecture for Extensible Self-Service Spoken Language Understanding
2,017
http://arxiv.org/pdf/1711.00549v4
Title ASK Building Architecture Extensible SelfService Spoken Language Understanding Summary paper present design machine learning architecture underlies Alexa Skills Kit ASK large scale Spoken Language Understanding SLU Software Development Kit SDK enables developer extend capability Amazons virtual assistant Alexa Am...
[0.049611739814281464, 0.004777115304023027, -0.041534118354320526, 0.024687061086297035, 0.0015271311858668923, -0.010411644354462624, 0.041570961475372314, -0.04209189862012863, 0.027954082936048508, -0.04792386293411255, 0.006490337662398815, 0.011914246715605259, -0.004317923448979855, 0.059516340494155884, 0.00837...
487
487
['Tomáš Kočiský', 'Jonathan Schwarz', 'Phil Blunsom', 'Chris Dyer', 'Karl Moritz Hermann', 'Gábor Melis', 'Edward Grefenstette']
1712.07040v1
Reading comprehension (RC)---in contrast to information retrieval---requires integrating information and reasoning about events, entities, and their relations across a full document. Question answering is conventionally used to assess RC ability, in both artificial agents and children learning to read. However, existin...
The NarrativeQA Reading Comprehension Challenge
2,017
http://arxiv.org/pdf/1712.07040v1
Title NarrativeQA Reading Comprehension Challenge Summary Reading comprehension RCin contrast information retrievalrequires integrating information reasoning event entity relation across full document Question answering conventionally used ass RC ability artificial agent child learning read However existing RC datasets...
[0.0365140326321125, 0.02348371408879757, -0.015957601368427277, 0.04249544069170952, -0.04983700439333916, 0.03662506863474846, 0.02855737879872322, -0.0033085576724261045, -0.004678906407207251, -0.010802393779158592, 0.0009602278587408364, 0.002951742149889469, 0.01648826152086258, 0.07350882887840271, -0.0335029475...
488
488
['Rajesh Bordawekar', 'Bortik Bandyopadhyay', 'Oded Shmueli']
1712.07199v1
We propose Cognitive Databases, an approach for transparently enabling Artificial Intelligence (AI) capabilities in relational databases. A novel aspect of our design is to first view the structured data source as meaningful unstructured text, and then use the text to build an unsupervised neural network model using a ...
Cognitive Database: A Step towards Endowing Relational Databases with Artificial Intelligence Capabilities
2,017
http://arxiv.org/pdf/1712.07199v1
Title Cognitive Database Step towards Endowing Relational Databases Artificial Intelligence Capabilities Summary propose Cognitive Databases approach transparently enabling Artificial Intelligence AI capability relational database novel aspect design first view structured data source meaningful unstructured text use te...
[0.02193552628159523, 0.0656096488237381, -0.026182014495134354, 0.01106394175440073, -0.025526393204927444, 0.019003618508577347, -0.006873030215501785, 0.016169670969247818, 0.07242733240127563, -0.02251298353075981, 0.01099012978374958, 0.023311303928494453, -0.032131098210811615, 0.07732986658811569, 0.032609082758...
489
489
['Iñigo Casanueva', 'Paweł Budzianowski', 'Pei-Hao Su', 'Stefan Ultes', 'Lina Rojas-Barahona', 'Bo-Hsiang Tseng', 'Milica Gašić']
1803.03232v1
Reinforcement learning (RL) is a promising approach to solve dialogue policy optimisation. Traditional RL algorithms, however, fail to scale to large domains due to the curse of dimensionality. We propose a novel Dialogue Management architecture, based on Feudal RL, which decomposes the decision into two steps; a first...
Feudal Reinforcement Learning for Dialogue Management in Large Domains
2,018
http://arxiv.org/pdf/1803.03232v1
Title Feudal Reinforcement Learning Dialogue Management Large Domains Summary Reinforcement learning RL promising approach solve dialogue policy optimisation Traditional RL algorithm however fail scale large domain due curse dimensionality propose novel Dialogue Management architecture based Feudal RL decomposes decisi...
[0.06175444647669792, 0.07851953059434891, -0.017721027135849, 0.0245374646037817, -0.022221723571419716, -0.0063163479790091515, 0.02900431491434574, -0.017272096127271652, -0.010333647951483727, -0.024914268404245377, -0.043090615421533585, -0.013307959772646427, -0.036684438586235046, 0.05240204185247421, -0.0522249...
490
490
['Stephen Merity', 'Nitish Shirish Keskar', 'Richard Socher']
1803.08240v1
Many of the leading approaches in language modeling introduce novel, complex and specialized architectures. We take existing state-of-the-art word level language models based on LSTMs and QRNNs and extend them to both larger vocabularies as well as character-level granularity. When properly tuned, LSTMs and QRNNs achie...
An Analysis of Neural Language Modeling at Multiple Scales
2,018
http://arxiv.org/pdf/1803.08240v1
Title Analysis Neural Language Modeling Multiple Scales Summary Many leading approach language modeling introduce novel complex specialized architecture take existing stateoftheart word level language model based LSTMs QRNNs extend larger vocabulary well characterlevel granularity properly tuned LSTMs QRNNs achieve sta...
[0.029181942343711853, 0.03138417750597, -0.012979317456483841, 0.04047025367617607, -0.006074283737689257, 0.014389488846063614, 0.03928348794579506, 0.02027115412056446, -0.018421638756990433, -0.015293341130018234, -0.03649844974279404, -0.08334321528673172, 0.03409227356314659, 0.0900942012667656, 0.012963213957846...
491
491
['Andreas Schwarz', 'Christian Huemmer', 'Roland Maas', 'Walter Kellermann']
1410.2479v2
We propose a spatial diffuseness feature for deep neural network (DNN)-based automatic speech recognition to improve recognition accuracy in reverberant and noisy environments. The feature is computed in real-time from multiple microphone signals without requiring knowledge or estimation of the direction of arrival, an...
Spatial Diffuseness Features for DNN-Based Speech Recognition in Noisy and Reverberant Environments
2,014
http://arxiv.org/pdf/1410.2479v2
Title Spatial Diffuseness Features DNNBased Speech Recognition Noisy Reverberant Environments Summary propose spatial diffuseness feature deep neural network DNNbased automatic speech recognition improve recognition accuracy reverberant noisy environment feature computed realtime multiple microphone signal without requ...
[-0.03159436956048012, 0.041587281972169876, 0.01891152933239937, 0.022850507870316505, -0.021994175389409065, -0.01445208303630352, 0.03328992426395416, -0.02419726923108101, -0.05811168625950813, 0.009771770797669888, -0.06108737364411354, 0.005841633304953575, 0.042951326817274094, -0.03019733354449272, 0.0137249864...
492
492
['Yoon Kim', 'Yacine Jernite', 'David Sontag', 'Alexander M. Rush']
1508.06615v4
We describe a simple neural language model that relies only on character-level inputs. Predictions are still made at the word-level. Our model employs a convolutional neural network (CNN) and a highway network over characters, whose output is given to a long short-term memory (LSTM) recurrent neural network language mo...
Character-Aware Neural Language Models
2,015
http://arxiv.org/pdf/1508.06615v4
Title CharacterAware Neural Language Models Summary describe simple neural language model relies characterlevel input Predictions still made wordlevel model employ convolutional neural network CNN highway network character whose output given long shortterm memory LSTM recurrent neural network language model RNNLM Engli...
[0.05119018629193306, 0.09103494882583618, -0.0036308716516941786, 0.0491948276758194, -0.012303113006055355, 0.0032413697335869074, 0.03976443409919739, 0.03735227882862091, -0.06278343498706818, -0.028260039165616035, -0.0029754447750747204, -0.10466524213552475, 0.058151911944150925, 0.05166538804769516, 0.058905340...
493
493
['Krzysztof Wołk', 'Krzysztof Marasek']
1509.08644v1
The quality of machine translation is rapidly evolving. Today one can find several machine translation systems on the web that provide reasonable translations, although the systems are not perfect. In some specific domains, the quality may decrease. A recently proposed approach to this domain is neural machine translat...
Neural-based machine translation for medical text domain. Based on European Medicines Agency leaflet texts
2,015
http://arxiv.org/pdf/1509.08644v1
Title Neuralbased machine translation medical text domain Based European Medicines Agency leaflet text Summary quality machine translation rapidly evolving Today one find several machine translation system web provide reasonable translation although system perfect specific domain quality may decrease recently proposed ...
[0.04181399196386337, -0.018864141777157784, 0.005011287517845631, 0.016684800386428833, -0.025218678638339043, 0.023490937426686287, 0.034301429986953735, 0.0345165841281414, -0.03873180225491524, -0.013261262327432632, 0.016608713194727898, -0.04808464273810387, 0.05090376362204552, 0.018950333818793297, 0.0345067940...
494
494
['Tsung-Hsien Wen', 'Milica Gasic', 'Nikola Mrksic', 'Lina M. Rojas-Barahona', 'Pei-Hao Su', 'Stefan Ultes', 'David Vandyke', 'Steve Young']
1606.03352v1
Recently a variety of LSTM-based conditional language models (LM) have been applied across a range of language generation tasks. In this work we study various model architectures and different ways to represent and aggregate the source information in an end-to-end neural dialogue system framework. A method called snaps...
Conditional Generation and Snapshot Learning in Neural Dialogue Systems
2,016
http://arxiv.org/pdf/1606.03352v1
Title Conditional Generation Snapshot Learning Neural Dialogue Systems Summary Recently variety LSTMbased conditional language model LM applied across range language generation task work study various model architecture different way represent aggregate source information endtoend neural dialogue system framework metho...
[0.04596489667892456, 0.041274070739746094, 0.00603374931961298, 0.029419537633657455, -0.02946520410478115, -0.009778846055269241, 0.04388587176799774, 0.012670128606259823, 0.017349235713481903, -0.04972049221396446, 0.004751638974994421, -0.061895038932561874, -0.008012899197638035, 0.05074413865804672, 0.0026315236...
495
495
['Julien Perez', 'Fei Liu']
1606.04052v5
In an end-to-end dialog system, the aim of dialog state tracking is to accurately estimate a compact representation of the current dialog status from a sequence of noisy observations produced by the speech recognition and the natural language understanding modules. This paper introduces a novel method of dialog state t...
Dialog state tracking, a machine reading approach using Memory Network
2,016
http://arxiv.org/pdf/1606.04052v5
Title Dialog state tracking machine reading approach using Memory Network Summary endtoend dialog system aim dialog state tracking accurately estimate compact representation current dialog status sequence noisy observation produced speech recognition natural language understanding module paper introduces novel method d...
[0.01852702721953392, 0.059703826904296875, -0.005396231543272734, 0.07672544568777084, -0.0025937859900295734, 0.010729890316724777, -0.011404688470065594, -0.0321563258767128, 0.0547000952064991, -0.051784448325634, 0.01922069676220417, -0.050492946058511734, -0.020785711705684662, 0.10536794364452362, -0.00875034835...
496
496
['Sándor Darányi', 'Peter Wittek', 'Konstantinos Konstantinidis', 'Symeon Papadopoulos', 'Efstratios Kontopoulos']
1608.01298v1
In accessibility tests for digital preservation, over time we experience drifts of localized and labelled content in statistical models of evolving semantics represented as a vector field. This articulates the need to detect, measure, interpret and model outcomes of knowledge dynamics. To this end we employ a high-perf...
A Physical Metaphor to Study Semantic Drift
2,016
http://arxiv.org/pdf/1608.01298v1
Title Physical Metaphor Study Semantic Drift Summary accessibility test digital preservation time experience drift localized labelled content statistical model evolving semantics represented vector field articulates need detect measure interpret model outcome knowledge dynamic end employ highperformance machine learnin...
[0.051582202315330505, 0.01668226160109043, -0.030742881819605827, 0.01801236905157566, -0.024954671040177345, -0.012800347059965134, 0.025951949879527092, 0.013821380212903023, -0.045094944536685944, -0.032826557755470276, 0.028548436239361763, 0.020404718816280365, 0.018836017698049545, 0.004220754373818636, 0.009514...
497
497
['Franck Dernoncourt', 'Ji Young Lee']
1609.08703v1
Systems based on artificial neural networks (ANNs) have achieved state-of-the-art results in many natural language processing tasks. Although ANNs do not require manually engineered features, ANNs have many hyperparameters to be optimized. The choice of hyperparameters significantly impacts models' performances. Howeve...
Optimizing Neural Network Hyperparameters with Gaussian Processes for Dialog Act Classification
2,016
http://arxiv.org/pdf/1609.08703v1
Title Optimizing Neural Network Hyperparameters Gaussian Processes Dialog Act Classification Summary Systems based artificial neural network ANNs achieved stateoftheart result many natural language processing task Although ANNs require manually engineered feature ANNs many hyperparameters optimized choice hyperparamete...
[0.027815915644168854, 0.043047185987234116, -0.011858375743031502, 0.0192045196890831, -0.01971953921020031, 0.0035111058969050646, 0.01757357455790043, 0.0236978679895401, 0.011024895124137402, -0.06809913367033005, -0.024109693244099617, 0.021360300481319427, 0.027190634980797768, 0.07266440242528915, -0.01505499333...
498
498
['Hans Krupakar', 'Keerthika Rajvel', 'Bharathi B', 'Angel Deborah S', 'Vallidevi Krishnamurthy']
1610.03934v1
Speech Translation has always been about giving source text or audio input and waiting for system to give translated output in desired form. In this paper, we present the Acoustic Dialect Decoder (ADD) - a voice to voice ear-piece translation device. We introduce and survey the recent advances made in the field of Spee...
A Survey of Voice Translation Methodologies - Acoustic Dialect Decoder
2,016
http://arxiv.org/pdf/1610.03934v1
Title Survey Voice Translation Methodologies Acoustic Dialect Decoder Summary Speech Translation always giving source text audio input waiting system give translated output desired form paper present Acoustic Dialect Decoder ADD voice voice earpiece translation device introduce survey recent advance made field Speech E...
[0.02819700725376606, -0.0026545983273535967, 0.007242600433528423, 0.019687477499246597, -0.040881067514419556, -0.018122613430023193, 0.034728530794382095, 0.007429505232721567, -0.03262017294764519, -0.02146136946976185, -0.03796542435884476, -0.008969255723059177, 0.08321838080883026, 0.0006240934017114341, 0.00728...
499
499
['Mark Neumann', 'Pontus Stenetorp', 'Sebastian Riedel']
1610.07647v2
Multi-hop inference is necessary for machine learning systems to successfully solve tasks such as Recognising Textual Entailment and Machine Reading. In this work, we demonstrate the effectiveness of adaptive computation for learning the number of inference steps required for examples of different complexity and that l...
Learning to Reason With Adaptive Computation
2,016
http://arxiv.org/pdf/1610.07647v2
Title Learning Reason Adaptive Computation Summary Multihop inference necessary machine learning system successfully solve task Recognising Textual Entailment Machine Reading work demonstrate effectiveness adaptive computation learning number inference step required example different complexity learning correct number ...
[0.04260840639472008, 0.00564073771238327, -0.030513713136315346, 0.040240053087472916, -0.06597141921520233, 0.028324350714683533, 0.013550831936299801, 0.02068973146378994, 0.01817084476351738, -0.041810546070337296, 0.07300622016191483, 0.061497610062360764, 0.02069713920354843, 0.046717461198568344, -0.010532931424...