Unnamed: 0.1
int64
0
41k
Unnamed: 0
int64
0
41k
author
stringlengths
9
1.39k
id
stringlengths
11
18
summary
stringlengths
25
3.66k
title
stringlengths
4
258
year
int64
1.99k
2.02k
arxiv_url
stringlengths
32
39
info
stringlengths
523
3.18k
embeddings
stringlengths
16.9k
17.1k
1,700
1,700
['Sarath Chandar A P', 'Stanislas Lauly', 'Hugo Larochelle', 'Mitesh M. Khapra', 'Balaraman Ravindran', 'Vikas Raykar', 'Amrita Saha']
1402.1454v1
Cross-language learning allows us to use training data from one language to build models for a different language. Many approaches to bilingual learning require that we have word-level alignment of sentences from parallel corpora. In this work we explore the use of autoencoder-based methods for cross-language learning ...
An Autoencoder Approach to Learning Bilingual Word Representations
2,014
http://arxiv.org/pdf/1402.1454v1
Title Autoencoder Approach Learning Bilingual Word Representations Summary Crosslanguage learning allows u use training data one language build model different language Many approach bilingual learning require wordlevel alignment sentence parallel corpus work explore use autoencoderbased method crosslanguage learning v...
[-0.019968630746006966, 0.02053377591073513, -0.02332456223666668, 0.08192688226699829, -0.049549203366041183, 0.022667380049824715, 0.028096262365579605, -0.0466051921248436, -0.0021242324728518724, -0.04964306578040123, -0.06189528852701187, -0.01866333745419979, 0.042168524116277695, -0.01780618354678154, -0.0095703...
1,701
1,701
['Yoav Goldberg', 'Omer Levy']
1402.3722v1
The word2vec software of Tomas Mikolov and colleagues (https://code.google.com/p/word2vec/ ) has gained a lot of traction lately, and provides state-of-the-art word embeddings. The learning models behind the software are described in two research papers. We found the description of the models in these papers to be some...
word2vec Explained: deriving Mikolov et al.'s negative-sampling word-embedding method
2,014
http://arxiv.org/pdf/1402.3722v1
Title word2vec Explained deriving Mikolov et al negativesampling wordembedding method Summary word2vec software Tomas Mikolov colleague httpscodegooglecompword2vec gained lot traction lately provides stateoftheart word embeddings learning model behind software described two research paper found description model paper ...
[0.04439600929617882, 0.0003964554634876549, -0.007454170845448971, 0.023938816040754318, -0.03111075982451439, -0.0012378176907077432, -0.04824608564376831, 0.009667136706411839, -0.023958172649145126, -0.03818759322166443, -0.016674863174557686, -0.025991516187787056, 0.043760381639003754, 0.02827337011694908, 0.0609...
1,702
1,702
['Dalei Wu', 'Haiqing Wu']
1405.4599v1
The trimming scheme with a prefixed cutoff portion is known as a method of improving the robustness of statistical models such as multivariate Gaussian mixture models (MG- MMs) in small scale tests by alleviating the impacts of outliers. However, when this method is applied to real- world data, such as noisy speech pro...
Modelling Data Dispersion Degree in Automatic Robust Estimation for Multivariate Gaussian Mixture Models with an Application to Noisy Speech Processing
2,014
http://arxiv.org/pdf/1405.4599v1
Title Modelling Data Dispersion Degree Automatic Robust Estimation Multivariate Gaussian Mixture Models Application Noisy Speech Processing Summary trimming scheme prefixed cutoff portion known method improving robustness statistical model multivariate Gaussian mixture model MG MMs small scale test alleviating impact o...
[-0.02667979523539543, -0.013502057641744614, -0.017377981916069984, 0.02664620243012905, 0.00651778606697917, 0.0011670138919726014, 0.04048670083284378, 0.007588709704577923, -0.041822612285614014, 0.011362195014953613, 0.004173398483544588, -0.015521174296736717, 0.07477845996618271, -0.008504309691488743, -0.009281...
1,703
1,703
['Daniel Godfrey', 'Caley Johns', 'Carl Meyer', 'Shaina Race', 'Carol Sadek']
1408.5427v1
Cluster analysis is a field of data analysis that extracts underlying patterns in data. One application of cluster analysis is in text-mining, the analysis of large collections of text to find similarities between documents. We used a collection of about 30,000 tweets extracted from Twitter just before the World Cup st...
A Case Study in Text Mining: Interpreting Twitter Data From World Cup Tweets
2,014
http://arxiv.org/pdf/1408.5427v1
Title Case Study Text Mining Interpreting Twitter Data World Cup Tweets Summary Cluster analysis field data analysis extract underlying pattern data One application cluster analysis textmining analysis large collection text find similarity document used collection 30000 tweet extracted Twitter World Cup started common ...
[0.06495460867881775, -0.026246817782521248, -0.036271076649427414, 0.06331506371498108, -0.0419095978140831, 0.024135345593094826, -0.04708484187722206, 0.02267981693148613, -0.006516085937619209, -0.017658893018960953, 0.025996744632720947, 0.01655764691531658, 0.019983889535069466, 0.05669841915369034, -0.0141307730...
1,704
1,704
['Kathryn Baker', 'Michael Bloodgood', 'Bonnie J. Dorr', 'Chris Callison-Burch', 'Nathaniel W. Filardo', 'Christine Piatko', 'Lori Levin', 'Scott Miller']
1502.01682v1
This paper describes the resource- and system-building efforts of an eight-week Johns Hopkins University Human Language Technology Center of Excellence Summer Camp for Applied Language Exploration (SCALE-2009) on Semantically-Informed Machine Translation (SIMT). We describe a new modality/negation (MN) annotation schem...
Use of Modality and Negation in Semantically-Informed Syntactic MT
2,015
http://arxiv.org/pdf/1502.01682v1
Title Use Modality Negation SemanticallyInformed Syntactic MT Summary paper describes resource systembuilding effort eightweek Johns Hopkins University Human Language Technology Center Excellence Summer Camp Applied Language Exploration SCALE2009 SemanticallyInformed Machine Translation SIMT describe new modalitynegati...
[0.03359067812561989, 0.03063707798719406, -0.01667138747870922, 0.02656002715229988, -0.06040293350815773, 0.005137065891176462, 0.026200873777270317, 0.0005940183764323592, -0.03476638346910477, -0.06327654421329498, -0.026056278496980667, 0.038499072194099426, 0.013805345632135868, 0.011957881040871143, 0.0250817090...
1,705
1,705
['Sanjeev Arora', 'Yuanzhi Li', 'Yingyu Liang', 'Tengyu Ma', 'Andrej Risteski']
1502.03520v7
Semantic word embeddings represent the meaning of a word via a vector, and are created by diverse methods. Many use nonlinear operations on co-occurrence statistics, and have hand-tuned hyperparameters and reweighting methods. This paper proposes a new generative model, a dynamic version of the log-linear topic model...
RAND-WALK: A Latent Variable Model Approach to Word Embeddings
2,015
http://arxiv.org/pdf/1502.03520v7
Title RANDWALK Latent Variable Model Approach Word Embeddings Summary Semantic word embeddings represent meaning word via vector created diverse method Many use nonlinear operation cooccurrence statistic handtuned hyperparameters reweighting method paper proposes new generative model dynamic version loglinear topic mod...
[0.03891397640109062, 0.015600455924868584, -0.008656934835016727, 0.05596701428294182, -0.01327063050121069, 0.008011867292225361, -0.0165289007127285, -0.010219096206128597, -0.035777099430561066, -0.04499467834830284, 0.052335526794195175, 0.014760655350983143, 0.013019159436225891, 0.03994184359908104, 0.0200957842...
1,706
1,706
['David Belanger', 'Sham Kakade']
1502.04081v2
Low dimensional representations of words allow accurate NLP models to be trained on limited annotated data. While most representations ignore words' local context, a natural way to induce context-dependent representations is to perform inference in a probabilistic latent-variable sequence model. Given the recent succes...
A Linear Dynamical System Model for Text
2,015
http://arxiv.org/pdf/1502.04081v2
Title Linear Dynamical System Model Text Summary Low dimensional representation word allow accurate NLP model trained limited annotated data representation ignore word local context natural way induce contextdependent representation perform inference probabilistic latentvariable sequence model Given recent success cont...
[0.01773865707218647, -0.0048938412219285965, -0.023702869191765785, 0.057768888771533966, -0.013394258916378021, -0.010737964883446693, -0.0017099156975746155, -0.010005648247897625, -0.015371222048997879, -0.06324589997529984, 0.029507381841540337, 0.0016296271933242679, 0.015548699535429478, 0.08409687876701355, 0.0...
1,707
1,707
['Thiago Marzagão']
1502.06161v1
This paper uses natural language processing to create the first machine-coded democracy index, which I call Automated Democracy Scores (ADS). The ADS are based on 42 million news articles from 6,043 different sources and cover all independent countries in the 1993-2012 period. Unlike the democracy indices we have today...
Using NLP to measure democracy
2,015
http://arxiv.org/pdf/1502.06161v1
Title Using NLP measure democracy Summary paper us natural language processing create first machinecoded democracy index call Automated Democracy Scores ADS ADS based 42 million news article 6043 different source cover independent country 19932012 period Unlike democracy index today ADS replicable standard error small ...
[0.02590905874967575, 0.09133400768041611, -0.022098444402217865, -0.008659042418003082, -0.022006185725331306, 0.012014330364763737, 0.017655080184340477, -0.02211349457502365, 0.010828303173184395, -0.05698423460125923, 0.06848876923322678, 0.0009296647040173411, 0.0274654533714056, 0.07806014269590378, -0.0358797796...
1,708
1,708
['Dani Yogatama', 'Noah A. Smith']
1503.00693v1
When applying machine learning to problems in NLP, there are many choices to make about how to represent input texts. These choices can have a big effect on performance, but they are often uninteresting to researchers or practitioners who simply need a module that performs well. We propose an approach to optimizing ove...
Bayesian Optimization of Text Representations
2,015
http://arxiv.org/pdf/1503.00693v1
Title Bayesian Optimization Text Representations Summary applying machine learning problem NLP many choice make represent input text choice big effect performance often uninteresting researcher practitioner simply need module performs well propose approach optimizing space choice formulating problem global optimization...
[0.04710850119590759, 0.019445616751909256, -0.01989380270242691, 0.023382076993584633, -0.05196037143468857, -0.002865392714738846, 0.016535911709070206, 0.059226322919130325, -0.0037789279595017433, -0.09349021315574646, 0.017565373331308365, 0.03674563392996788, 0.015989704057574272, 0.07484100759029388, -0.02781843...
1,709
1,709
['Vinodkumar Prabhakaran', 'Michael Bloodgood', 'Mona Diab', 'Bonnie Dorr', 'Lori Levin', 'Christine D. Piatko', 'Owen Rambow', 'Benjamin Van Durme']
1503.01190v1
We explore training an automatic modality tagger. Modality is the attitude that a speaker might have toward an event or state. One of the main hurdles for training a linguistic tagger is gathering training data. This is particularly problematic for training a tagger for modality because modality triggers are sparse for...
Statistical modality tagging from rule-based annotations and crowdsourcing
2,015
http://arxiv.org/pdf/1503.01190v1
Title Statistical modality tagging rulebased annotation crowdsourcing Summary explore training automatic modality tagger Modality attitude speaker might toward event state One main hurdle training linguistic tagger gathering training data particularly problematic training tagger modality modality trigger sparse overwhe...
[0.044138189405202866, 0.05850784853100777, -0.016779422760009766, -0.008176017552614212, -0.023389138281345367, 0.005461566150188446, 0.0014529911568388343, -0.011455340310931206, -0.041785191744565964, -0.09863385558128357, 0.015791695564985275, 0.022585678845643997, 0.004246774595230818, 0.0330125130712986, -0.02424...
1,710
1,710
['Luke Vilnis', 'David Belanger', 'Daniel Sheldon', 'Andrew McCallum']
1503.01397v3
Many inference problems in structured prediction are naturally solved by augmenting a tractable dependency structure with complex, non-local auxiliary objectives. This includes the mean field family of variational inference algorithms, soft- or hard-constrained inference using Lagrangian relaxation or linear programmin...
Bethe Projections for Non-Local Inference
2,015
http://arxiv.org/pdf/1503.01397v3
Title Bethe Projections NonLocal Inference Summary Many inference problem structured prediction naturally solved augmenting tractable dependency structure complex nonlocal auxiliary objective includes mean field family variational inference algorithm soft hardconstrained inference using Lagrangian relaxation linear pro...
[-0.00933182891458273, 0.09850789606571198, -0.004147167317569256, 0.04936450719833374, -0.048886436969041824, 0.008956230245530605, 0.05803922936320305, 0.020104771479964256, -0.029476413503289223, -0.040510665625333786, -0.0054901293478906155, -0.005789319518953562, 0.05983858183026314, 0.00869346410036087, 0.0398116...
1,711
1,711
['Junyu Xuan', 'Jie Lu', 'Guangquan Zhang', 'Richard Yi Da Xu', 'Xiangfeng Luo']
1503.08542v1
Traditional Relational Topic Models provide a way to discover the hidden topics from a document network. Many theoretical and practical tasks, such as dimensional reduction, document clustering, link prediction, benefit from this revealed knowledge. However, existing relational topic models are based on an assumption t...
Nonparametric Relational Topic Models through Dependent Gamma Processes
2,015
http://arxiv.org/pdf/1503.08542v1
Title Nonparametric Relational Topic Models Dependent Gamma Processes Summary Traditional Relational Topic Models provide way discover hidden topic document network Many theoretical practical task dimensional reduction document clustering link prediction benefit revealed knowledge However existing relational topic mode...
[0.023288603872060776, 0.024098718538880348, -0.02897268533706665, 0.003010866930708289, -0.03300043195486069, -0.0003807863686233759, -0.001211307244375348, 0.012763189151883125, -0.013953671790659428, -0.029283445328474045, -0.049471884965896606, 0.030756717547774315, -0.005458103492856026, 0.030566420406103134, -0.0...
1,712
1,712
['Jiaji Huang', 'Rewon Child', 'Vinay Rao', 'Hairong Liu', 'Sanjeev Satheesh', 'Adam Coates']
1612.03226v1
In training speech recognition systems, labeling audio clips can be expensive, and not all data is equally valuable. Active learning aims to label only the most informative samples to reduce cost. For speech recognition, confidence scores and other likelihood-based active learning methods have been shown to be effectiv...
Active Learning for Speech Recognition: the Power of Gradients
2,016
http://arxiv.org/pdf/1612.03226v1
Title Active Learning Speech Recognition Power Gradients Summary training speech recognition system labeling audio clip expensive data equally valuable Active learning aim label informative sample reduce cost speech recognition confidence score likelihoodbased active learning method shown effective Gradientbased active...
[-0.005078149493783712, -0.023373359814286232, 0.001576327602379024, -0.009753570891916752, 0.01638173498213291, 0.004150940105319023, 0.04586920887231827, -0.044032126665115356, -0.035827428102493286, -0.0032846671529114246, -0.03097865730524063, -0.002329138806089759, 0.05617901682853699, 0.025111068040132523, 0.0269...
1,713
1,713
['Eric S. Tellez', 'Sabino Miranda Jiménez', 'Mario Graff', 'Daniela Moctezuma', 'Ranyart R. Suárez', 'Oscar S. Siordia']
1612.05270v1
Recently, sentiment analysis has received a lot of attention due to the interest in mining opinions of social media users. Sentiment analysis consists in determining the polarity of a given text, i.e., its degree of positiveness or negativeness. Traditionally, Sentiment Analysis algorithms have been tailored to a speci...
A Simple Approach to Multilingual Polarity Classification in Twitter
2,016
http://arxiv.org/pdf/1612.05270v1
Title Simple Approach Multilingual Polarity Classification Twitter Summary Recently sentiment analysis received lot attention due interest mining opinion social medium user Sentiment analysis consists determining polarity given text ie degree positiveness negativeness Traditionally Sentiment Analysis algorithm tailored...
[0.00213824724778533, 0.03926084190607071, -0.01181692723184824, 0.05207912623882294, -0.042175695300102234, 0.031833577901124954, -0.02503800392150879, -0.01682036928832531, 0.011418795213103294, -0.04893822595477104, -0.021675050258636475, -0.022530118003487587, 0.0029864246025681496, 0.03417373448610306, -0.02542760...
1,714
1,714
['Leila Arras', 'Franziska Horn', 'Grégoire Montavon', 'Klaus-Robert Müller', 'Wojciech Samek']
1612.07843v1
Text documents can be described by a number of abstract concepts such as semantic category, writing style, or sentiment. Machine learning (ML) models have been trained to automatically map documents to these abstract concepts, allowing to annotate very large text collections, more than could be processed by a human in ...
"What is Relevant in a Text Document?": An Interpretable Machine Learning Approach
2,016
http://arxiv.org/pdf/1612.07843v1
Title Relevant Text Document Interpretable Machine Learning Approach Summary Text document described number abstract concept semantic category writing style sentiment Machine learning ML model trained automatically map document abstract concept allowing annotate large text collection could processed human lifetime Besi...
[0.06186361238360405, 0.0324312187731266, 0.012833287939429283, 0.02312200516462326, -0.0383351631462574, 0.01817934587597847, 0.001925541670061648, 0.039196792989969254, -0.016549192368984222, -0.08214789628982544, -0.014797106385231018, 0.04710216075181961, -0.00925437081605196, 0.03795032575726509, -0.03087709099054...
1,715
1,715
['Jun Zhu', 'Xun Zheng', 'Bo Zhang']
1310.2408v1
Supervised topic models with a logistic likelihood have two issues that potentially limit their practical use: 1) response variables are usually over-weighted by document word counts; and 2) existing variational inference methods make strict mean-field assumptions. We address these issues by: 1) introducing a regulariz...
Improved Bayesian Logistic Supervised Topic Models with Data Augmentation
2,013
http://arxiv.org/pdf/1310.2408v1
Title Improved Bayesian Logistic Supervised Topic Models Data Augmentation Summary Supervised topic model logistic likelihood two issue potentially limit practical use 1 response variable usually overweighted document word count 2 existing variational inference method make strict meanfield assumption address issue 1 in...
[0.060598380863666534, 0.02794906310737133, -0.010377953760325909, -0.007321027107536793, -0.009390396997332573, -0.00015138353046495467, 0.011987539008259773, -0.006888849660754204, -0.04430720955133438, -0.05959027260541916, 0.009445952251553535, 0.028049945831298828, -0.012185932137072086, 0.03736753761768341, -0.01...
1,716
1,716
['Roland Maas', 'Christian Huemmer', 'Armin Sehr', 'Walter Kellermann']
1310.3099v2
This article provides a unifying Bayesian network view on various approaches for acoustic model adaptation, missing feature, and uncertainty decoding that are well-known in the literature of robust automatic speech recognition. The representatives of these classes can often be deduced from a Bayesian network that exten...
A Bayesian Network View on Acoustic Model-Based Techniques for Robust Speech Recognition
2,013
http://arxiv.org/pdf/1310.3099v2
Title Bayesian Network View Acoustic ModelBased Techniques Robust Speech Recognition Summary article provides unifying Bayesian network view various approach acoustic model adaptation missing feature uncertainty decoding wellknown literature robust automatic speech recognition representative class often deduced Bayesia...
[0.00019893849093932658, 0.030020350590348244, 0.0037613818421959877, 0.022657057270407677, -0.007741710636764765, -0.00500610563904047, 0.028634047135710716, 0.008041816763579845, -0.058472324162721634, -0.02899346314370632, -0.008224532008171082, 0.007312949746847153, 0.06827350705862045, 0.04346977546811104, 0.00062...
1,717
1,717
['Tomas Mikolov', 'Ilya Sutskever', 'Kai Chen', 'Greg Corrado', 'Jeffrey Dean']
1310.4546v1
The recently introduced continuous Skip-gram model is an efficient method for learning high-quality distributed vector representations that capture a large number of precise syntactic and semantic word relationships. In this paper we present several extensions that improve both the quality of the vectors and the traini...
Distributed Representations of Words and Phrases and their Compositionality
2,013
http://arxiv.org/pdf/1310.4546v1
Title Distributed Representations Words Phrases Compositionality Summary recently introduced continuous Skipgram model efficient method learning highquality distributed vector representation capture large number precise syntactic semantic word relationship paper present several extension improve quality vector training...
[0.04475008696317673, 0.011088273487985134, -0.013445473276078701, 0.04694761708378792, -0.027753349393606186, 0.007069509010761976, -0.017947211861610413, -0.006088783498853445, -0.0270652137696743, -0.060532428324222565, -0.011708070524036884, -0.012277013622224331, 0.026394668966531754, 0.026513494551181793, 0.01510...
1,718
1,718
['Ozan İrsoy', 'Claire Cardie']
1312.0493v1
Recently, deep architectures, such as recurrent and recursive neural networks have been successfully applied to various natural language processing tasks. Inspired by bidirectional recurrent neural networks which use representations that summarize the past and future around an instance, we propose a novel architecture ...
Bidirectional Recursive Neural Networks for Token-Level Labeling with Structure
2,013
http://arxiv.org/pdf/1312.0493v1
Title Bidirectional Recursive Neural Networks TokenLevel Labeling Structure Summary Recently deep architecture recurrent recursive neural network successfully applied various natural language processing task Inspired bidirectional recurrent neural network use representation summarize past future around instance propose...
[0.05301997438073158, 0.020764442160725594, 0.012574116699397564, 0.06228034570813179, -0.04722466319799423, -0.0031344559974968433, -0.018177947029471397, -0.020312506705522537, -0.020932044833898544, -0.06638488918542862, 0.003838426200672984, -0.011009251698851585, -0.016119299456477165, 0.09089713543653488, -0.0015...
1,719
1,719
['Ankur P. Parikh', 'Avneesh Saluja', 'Chris Dyer', 'Eric P. Xing']
1312.7077v2
We present power low rank ensembles (PLRE), a flexible framework for n-gram language modeling where ensembles of low rank matrices and tensors are used to obtain smoothed probability estimates of words in context. Our method can be understood as a generalization of n-gram modeling to non-integer n, and includes standar...
Language Modeling with Power Low Rank Ensembles
2,013
http://arxiv.org/pdf/1312.7077v2
Title Language Modeling Power Low Rank Ensembles Summary present power low rank ensemble PLRE flexible framework ngram language modeling ensemble low rank matrix tensor used obtain smoothed probability estimate word context method understood generalization ngram modeling noninteger n includes standard technique absolut...
[0.030020473524928093, 0.03275823965668678, -0.022202201187610626, 0.0262181106954813, -0.038318805396556854, 0.025147980079054832, 0.0020431354641914368, 0.003641626564785838, -0.04061627388000488, -0.04393815994262695, 0.0018784194253385067, -0.07516767829656601, 0.06765775382518768, -0.007486241869628429, 0.05242437...
1,720
1,720
['Michael Bloodgood', 'K. Vijay-Shanker']
1409.3881v1
There is a broad range of BioNLP tasks for which active learning (AL) can significantly reduce annotation costs and a specific AL algorithm we have developed is particularly effective in reducing annotation costs for these tasks. We have previously developed an AL algorithm called ClosestInitPA that works best with tas...
An Approach to Reducing Annotation Costs for BioNLP
2,014
http://arxiv.org/pdf/1409.3881v1
Title Approach Reducing Annotation Costs BioNLP Summary broad range BioNLP task active learning AL significantly reduce annotation cost specific AL algorithm developed particularly effective reducing annotation cost task previously developed AL algorithm called ClosestInitPA work best task following characteristic redu...
[0.033342402428388596, -0.00923463050276041, -0.006645193789154291, -0.025067854672670364, -0.009983290918171406, 0.015600084327161312, 0.016952509060502052, 0.05016056075692177, 0.006922307889908552, -0.06216815114021301, 0.0037310386542230844, 0.04750100150704384, 0.015300319530069828, 0.013908120803534985, 0.0135505...
1,721
1,721
['Michael Bloodgood', 'K. Vijay-Shanker']
1409.4835v1
Actively sampled data can have very different characteristics than passively sampled data. Therefore, it's promising to investigate using different inference procedures during AL than are used during passive learning (PL). This general idea is explored in detail for the focused case of AL with cost-weighted SVMs for im...
Taking into Account the Differences between Actively and Passively Acquired Data: The Case of Active Learning with Support Vector Machines for Imbalanced Datasets
2,014
http://arxiv.org/pdf/1409.4835v1
Title Taking Account Differences Actively Passively Acquired Data Case Active Learning Support Vector Machines Imbalanced Datasets Summary Actively sampled data different characteristic passively sampled data Therefore promising investigate using different inference procedure AL used passive learning PL general idea ex...
[0.026493163779377937, -0.05138388276100159, -0.04536108300089836, -0.011343173682689667, 0.02367592044174671, 0.037705812603235245, 0.010170185007154942, -0.00599520280957222, -0.016301658004522324, -0.055204279720783234, 0.016717346385121346, 0.013962133787572384, 0.03867000713944435, 0.022816786542534828, 0.00996683...
1,722
1,722
['Michael Bloodgood', 'K. Vijay-Shanker']
1409.5165v1
A survey of existing methods for stopping active learning (AL) reveals the needs for methods that are: more widely applicable; more aggressive in saving annotations; and more stable across changing datasets. A new method for stopping AL based on stabilizing predictions is presented that addresses these needs. Furthermo...
A Method for Stopping Active Learning Based on Stabilizing Predictions and the Need for User-Adjustable Stopping
2,014
http://arxiv.org/pdf/1409.5165v1
Title Method Stopping Active Learning Based Stabilizing Predictions Need UserAdjustable Stopping Summary survey existing method stopping active learning AL reveals need method widely applicable aggressive saving annotation stable across changing datasets new method stopping AL based stabilizing prediction presented add...
[-0.03487914428114891, 0.046756964176893234, -0.02695629559457302, -0.03548338636755943, 0.05479619279503822, 0.013501619920134544, 0.02504039742052555, -0.007747399155050516, -0.0235403124243021, -0.012751717120409012, 0.05754479020833969, 0.007323394529521465, -0.021816875785589218, 0.07684693485498428, -0.0272496081...
1,723
1,723
['Kathryn Baker', 'Michael Bloodgood', 'Chris Callison-Burch', 'Bonnie J. Dorr', 'Nathaniel W. Filardo', 'Lori Levin', 'Scott Miller', 'Christine Piatko']
1409.7085v1
We describe a unified and coherent syntactic framework for supporting a semantically-informed syntactic approach to statistical machine translation. Semantically enriched syntactic tags assigned to the target-language training texts improved translation quality. The resulting system significantly outperformed a linguis...
Semantically-Informed Syntactic Machine Translation: A Tree-Grafting Approach
2,014
http://arxiv.org/pdf/1409.7085v1
Title SemanticallyInformed Syntactic Machine Translation TreeGrafting Approach Summary describe unified coherent syntactic framework supporting semanticallyinformed syntactic approach statistical machine translation Semantically enriched syntactic tag assigned targetlanguage training text improved translation quality r...
[0.04547286033630371, -0.0019383601611480117, -0.01930360496044159, 0.045233312994241714, -0.0648532509803772, 0.023688172921538353, -0.01139404159039259, -0.0022135861217975616, -0.04720698669552803, -0.046445295214653015, 0.006703452207148075, 0.007922899909317493, 0.04489854350686073, -0.028527043759822845, 0.031413...
1,724
1,724
['Stephan Gouws', 'Yoshua Bengio', 'Greg Corrado']
1410.2455v3
We introduce BilBOWA (Bilingual Bag-of-Words without Alignments), a simple and computationally-efficient model for learning bilingual distributed representations of words which can scale to large monolingual datasets and does not require word-aligned parallel training data. Instead it trains directly on monolingual dat...
BilBOWA: Fast Bilingual Distributed Representations without Word Alignments
2,014
http://arxiv.org/pdf/1410.2455v3
Title BilBOWA Fast Bilingual Distributed Representations without Word Alignments Summary introduce BilBOWA Bilingual BagofWords without Alignments simple computationallyefficient model learning bilingual distributed representation word scale large monolingual datasets require wordaligned parallel training data Instead ...
[-0.006022888235747814, 0.03113514557480812, -0.0027029297780245543, 0.0884915217757225, -0.039717916399240494, 0.042330943048000336, 0.05094214156270027, -0.03845565766096115, 0.020851818844676018, -0.021599335595965385, -0.060442693531513214, -0.0524483285844326, 0.034306593239307404, -0.03652220591902733, 0.00561730...
1,725
1,725
['Finale Doshi-Velez', 'Byron Wallace', 'Ryan Adams']
1410.4510v2
Originally designed to model text, topic modeling has become a powerful tool for uncovering latent structure in domains including medicine, finance, and vision. The goals for the model vary depending on the application: in some cases, the discovered topics may be used for prediction or some other downstream task. In ot...
Graph-Sparse LDA: A Topic Model with Structured Sparsity
2,014
http://arxiv.org/pdf/1410.4510v2
Title GraphSparse LDA Topic Model Structured Sparsity Summary Originally designed model text topic modeling become powerful tool uncovering latent structure domain including medicine finance vision goal model vary depending application case discovered topic may used prediction downstream task case content topic may int...
[0.014826451428234577, 0.03852814808487892, -0.01338238175958395, 0.014699302613735199, 7.172942423494533e-05, -0.004249191842973232, -0.0026000619400292635, 0.05552804097533226, -0.050952520221471786, -0.017888512462377548, 0.030601125210523605, 0.0065007079392671585, 0.02280432917177677, 0.09721865504980087, 0.002779...
1,726
1,726
['Michael Bloodgood', 'Chris Callison-Burch']
1410.5491v1
Building machine translation (MT) test sets is a relatively expensive task. As MT becomes increasingly desired for more and more language pairs and more and more domains, it becomes necessary to build test sets for each case. In this paper, we investigate using Amazon's Mechanical Turk (MTurk) to make MT test sets chea...
Using Mechanical Turk to Build Machine Translation Evaluation Sets
2,014
http://arxiv.org/pdf/1410.5491v1
Title Using Mechanical Turk Build Machine Translation Evaluation Sets Summary Building machine translation MT test set relatively expensive task MT becomes increasingly desired language pair domain becomes necessary build test set case paper investigate using Amazons Mechanical Turk MTurk make MT test set cheaply find ...
[0.03499825298786163, 0.026640072464942932, -0.04081592336297035, 0.024303654208779335, -0.04445013776421547, 0.004387443885207176, 0.07096207141876221, 0.025856032967567444, -0.06749677658081055, -0.043730754405260086, 0.005153421312570572, -0.022020194679498672, 0.037159264087677, 0.019903426989912987, 0.011455481871...
1,727
1,727
['Michael Bloodgood', 'Chris Callison-Burch']
1410.5877v1
We explore how to improve machine translation systems by adding more translation data in situations where we already have substantial resources. The main challenge is how to buck the trend of diminishing returns that is commonly encountered. We present an active learning-style data solicitation algorithm to meet this c...
Bucking the Trend: Large-Scale Cost-Focused Active Learning for Statistical Machine Translation
2,014
http://arxiv.org/pdf/1410.5877v1
Title Bucking Trend LargeScale CostFocused Active Learning Statistical Machine Translation Summary explore improve machine translation system adding translation data situation already substantial resource main challenge buck trend diminishing return commonly encountered present active learningstyle data solicitation al...
[0.010844266973435879, -0.013924607075750828, -0.021593287587165833, -0.007557649631053209, -0.02845119498670101, 0.0295419879257679, 0.03605559095740318, 0.014396809972822666, -0.0343991257250309, -0.04976285248994827, 0.015682218596339226, -0.008743313141167164, 0.029743656516075134, 0.05917927622795105, 0.0211931709...
1,728
1,728
['Michael Bloodgood', 'Peng Ye', 'Paul Rodrigues', 'David Zajic', 'David Doermann']
1410.8553v1
When digitizing a print bilingual dictionary, whether via optical character recognition or manual entry, it is inevitable that errors are introduced into the electronic version that is created. We investigate automating the process of detecting errors in an XML representation of a digitized print dictionary using a hyb...
A random forest system combination approach for error detection in digital dictionaries
2,014
http://arxiv.org/pdf/1410.8553v1
Title random forest system combination approach error detection digital dictionary Summary digitizing print bilingual dictionary whether via optical character recognition manual entry inevitable error introduced electronic version created investigate automating process detecting error XML representation digitized print...
[0.03034163825213909, 0.02360486052930355, -0.007509356830269098, 0.05264563485980034, -0.0043377503752708435, 0.019954616203904152, 0.0019734855741262436, 0.0429847352206707, 0.013488280586898327, -0.05592608451843262, 0.05320899933576584, 0.009796632453799248, 0.034290872514247894, 0.014869872480630875, -0.0553064011...
1,729
1,729
['John E. Miller', 'Michael Bloodgood', 'Manabu Torii', 'K. Vijay-Shanker']
1411.0007v1
Part-of-speech (POS) tagging is a fundamental component for performing natural language tasks such as parsing, information extraction, and question answering. When POS taggers are trained in one domain and applied in significantly different domains, their performance can degrade dramatically. We present a methodology f...
Rapid Adaptation of POS Tagging for Domain Specific Uses
2,014
http://arxiv.org/pdf/1411.0007v1
Title Rapid Adaptation POS Tagging Domain Specific Uses Summary Partofspeech POS tagging fundamental component performing natural language task parsing information extraction question answering POS tagger trained one domain applied significantly different domain performance degrade dramatically present methodology rapi...
[0.07653313130140305, 0.01328081265091896, 0.0069460500963032246, -0.037814829498529434, -0.02148306742310524, 0.017386239022016525, 0.0463210754096508, 0.039404865354299545, -0.016370238736271858, -0.08887553960084915, 0.0018233683658763766, 0.02948838844895363, 0.011210024356842041, 0.05438472330570221, -0.0015382926...
1,730
1,730
['Fangjian Guo', 'Charles Blundell', 'Hanna Wallach', 'Katherine Heller']
1411.2674v3
We present the Bayesian Echo Chamber, a new Bayesian generative model for social interaction data. By modeling the evolution of people's language usage over time, this model discovers latent influence relationships between them. Unlike previous work on inferring influence, which has primarily focused on simple temporal...
The Bayesian Echo Chamber: Modeling Social Influence via Linguistic Accommodation
2,014
http://arxiv.org/pdf/1411.2674v3
Title Bayesian Echo Chamber Modeling Social Influence via Linguistic Accommodation Summary present Bayesian Echo Chamber new Bayesian generative model social interaction data modeling evolution people language usage time model discovers latent influence relationship Unlike previous work inferring influence primarily fo...
[0.06180011108517647, 0.09379434585571289, -0.04919329658150673, -0.022013243287801743, -0.020863644778728485, 0.013265486806631088, 0.0447920560836792, 0.0009603506769053638, -0.038023754954338074, -0.035742200911045074, 0.030587514862418175, 0.005485658999532461, 0.036158446222543716, 0.03528127074241638, 0.019729437...
1,731
1,731
['Bishan Yang', 'Wen-tau Yih', 'Xiaodong He', 'Jianfeng Gao', 'Li Deng']
1411.4072v1
In this paper we present a unified framework for modeling multi-relational representations, scoring, and learning, and conduct an empirical study of several recent multi-relational embedding models under the framework. We investigate the different choices of relation operators based on linear and bilinear transformatio...
Learning Multi-Relational Semantics Using Neural-Embedding Models
2,014
http://arxiv.org/pdf/1411.4072v1
Title Learning MultiRelational Semantics Using NeuralEmbedding Models Summary paper present unified framework modeling multirelational representation scoring learning conduct empirical study several recent multirelational embedding model framework investigate different choice relation operator based linear bilinear tra...
[0.019642969593405724, 0.015632541850209236, -0.010109800845384598, 0.04931814596056938, 0.00559957092627883, 0.02199227176606655, -0.02310774475336075, -0.005205356981605291, 0.010075250640511513, -0.06854714453220367, -0.002090172842144966, 0.03963708505034447, -0.04428543150424957, -0.01682754047214985, 0.0188943929...
1,732
1,732
['Tianze Shi', 'Zhiyuan Liu']
1411.5595v2
The Global Vectors for word representation (GloVe), introduced by Jeffrey Pennington et al. is reported to be an efficient and effective method for learning vector representations of words. State-of-the-art performance is also provided by skip-gram with negative-sampling (SGNS) implemented in the word2vec tool. In this...
Linking GloVe with word2vec
2,014
http://arxiv.org/pdf/1411.5595v2
Title Linking GloVe word2vec Summary Global Vectors word representation GloVe introduced Jeffrey Pennington et al reported efficient effective method learning vector representation word Stateoftheart performance also provided skipgram negativesampling SGNS implemented word2vec tool note explain similarity training obje...
[0.020341457799077034, 0.005737768951803446, -0.00841280072927475, 0.0192551352083683, -0.02720467932522297, 0.014333861880004406, -0.010640564374625683, 0.04119884595274925, -0.009094865061342716, -0.04695333167910576, -0.011486828327178955, -0.022626396268606186, 0.03364413604140282, 0.038211215287446976, 0.036655984...
1,733
1,733
['Suleyman Cetintas', 'Luo Si', 'Yan Ping Xin', 'Dake Zhang', 'Joo Young Park', 'Ron Tzur']
1411.5732v1
Estimating the difficulty level of math word problems is an important task for many educational applications. Identification of relevant and irrelevant sentences in math word problems is an important step for calculating the difficulty levels of such problems. This paper addresses a novel application of text categoriza...
A Joint Probabilistic Classification Model of Relevant and Irrelevant Sentences in Mathematical Word Problems
2,014
http://arxiv.org/pdf/1411.5732v1
Title Joint Probabilistic Classification Model Relevant Irrelevant Sentences Mathematical Word Problems Summary Estimating difficulty level math word problem important task many educational application Identification relevant irrelevant sentence math word problem important step calculating difficulty level problem pape...
[0.04393162578344345, -7.79133551986888e-05, -0.02409650944173336, 0.021500522270798683, -0.04818441718816757, 0.0299435555934906, 0.02609424851834774, 0.016524983569979668, -0.00032894129981286824, -0.08584391325712204, 0.037934668362140656, 0.026072390377521515, 0.04511096701025963, 0.0014033288462087512, -0.04410924...
1,734
1,734
['Rie Johnson', 'Tong Zhang']
1412.1058v2
Convolutional neural network (CNN) is a neural network that can make use of the internal structure of data such as the 2D structure of image data. This paper studies CNN on text categorization to exploit the 1D structure (namely, word order) of text data for accurate prediction. Instead of using low-dimensional word ve...
Effective Use of Word Order for Text Categorization with Convolutional Neural Networks
2,014
http://arxiv.org/pdf/1412.1058v2
Title Effective Use Word Order Text Categorization Convolutional Neural Networks Summary Convolutional neural network CNN neural network make use internal structure data 2D structure image data paper study CNN text categorization exploit 1D structure namely word order text data accurate prediction Instead using lowdime...
[0.07230666279792786, 0.04432699829339981, 0.03302713483572006, 0.08804094046354294, -0.0206686370074749, -0.006297307088971138, 0.020050842314958572, 0.046456947922706604, 0.027751106768846512, -0.07065077126026154, -0.023469770327210426, 0.009037727490067482, 0.002495946129783988, 0.023086125031113625, 0.000921291997...
1,735
1,735
['Ivan Titov', 'Ehsan Khoddam']
1412.6418v3
In this work, we propose a new method to integrate two recent lines of work: unsupervised induction of shallow semantics (e.g., semantic roles) and factorization of relations in text and knowledge bases. Our model consists of two components: (1) an encoding component: a semantic role labeling model which predicts roles...
Inducing Semantic Representation from Text by Jointly Predicting and Factorizing Relations
2,014
http://arxiv.org/pdf/1412.6418v3
Title Inducing Semantic Representation Text Jointly Predicting Factorizing Relations Summary work propose new method integrate two recent line work unsupervised induction shallow semantics eg semantic role factorization relation text knowledge base model consists two component 1 encoding component semantic role labelin...
[0.042718809098005295, 0.013532069511711597, -0.001741364598274231, 0.06887433677911758, -0.013049871660768986, 0.018213259056210518, -0.018762163817882538, -0.01104635838419199, -0.022371040657162666, -0.0710696429014206, -0.024732839316129684, -0.005300404969602823, -0.014049042947590351, 0.04437480866909027, -0.0128...
1,736
1,736
['Ozan İrsoy', 'Claire Cardie']
1412.6577v3
We present the multiplicative recurrent neural network as a general model for compositional meaning in language, and evaluate it on the task of fine-grained sentiment analysis. We establish a connection to the previously investigated matrix-space models for compositionality, and show they are special cases of the multi...
Modeling Compositionality with Multiplicative Recurrent Neural Networks
2,014
http://arxiv.org/pdf/1412.6577v3
Title Modeling Compositionality Multiplicative Recurrent Neural Networks Summary present multiplicative recurrent neural network general model compositional meaning language evaluate task finegrained sentiment analysis establish connection previously investigated matrixspace model compositionality show special case mul...
[0.018285095691680908, 0.06698916107416153, 0.0032731089740991592, 0.043800778687000275, -0.02277706377208233, 0.021135712042450905, -0.029314570128917694, -0.03179441764950752, -0.06986979395151138, -0.036112040281295776, -0.010236089117825031, -0.04251963272690773, 0.0019263752037659287, 0.029431598260998726, -0.0047...
1,737
1,737
['Jinseok Nam', 'Johannes Fürnkranz']
1412.6881v3
An important problem in multi-label classification is to capture label patterns or underlying structures that have an impact on such patterns. This paper addresses one such problem, namely how to exploit hierarchical structures over labels. We present a novel method to learn vector representations of a label space give...
On Learning Vector Representations in Hierarchical Label Spaces
2,014
http://arxiv.org/pdf/1412.6881v3
Title Learning Vector Representations Hierarchical Label Spaces Summary important problem multilabel classification capture label pattern underlying structure impact pattern paper address one problem namely exploit hierarchical structure label present novel method learn vector representation label space given hierarchy...
[0.018613535910844803, 0.03689483180642128, -0.03261834755539894, 0.051269128918647766, 0.013546140864491463, 0.014273268170654774, 0.018884718418121338, 0.039152808487415314, 0.004139601718634367, -0.05067514628171921, -0.03347432613372803, 0.022753920406103134, -0.023351099342107773, -0.014156201854348183, -0.0076203...
1,738
1,738
['Oriol Vinyals', 'Lukasz Kaiser', 'Terry Koo', 'Slav Petrov', 'Ilya Sutskever', 'Geoffrey Hinton']
1412.7449v3
Syntactic constituency parsing is a fundamental problem in natural language processing and has been the subject of intensive research and engineering for decades. As a result, the most accurate parsers are domain specific, complex, and inefficient. In this paper we show that the domain agnostic attention-enhanced seque...
Grammar as a Foreign Language
2,014
http://arxiv.org/pdf/1412.7449v3
Title Grammar Foreign Language Summary Syntactic constituency parsing fundamental problem natural language processing subject intensive research engineering decade result accurate parser domain specific complex inefficient paper show domain agnostic attentionenhanced sequencetosequence model achieves stateoftheart resu...
[0.06937654316425323, 0.035618722438812256, 0.0004150472523178905, 0.047776516526937485, -0.04470133036375046, 0.014970286749303341, -0.016503838822245598, -0.04197145625948906, 0.005775857716798782, -0.08082158118486404, 0.009722680784761906, -0.04839170724153519, 0.011022642254829407, -0.0004761804302688688, 0.024709...
1,739
1,739
['Lars Maaloe', 'Morten Arngren', 'Ole Winther']
1501.04325v1
Applying traditional collaborative filtering to digital publishing is challenging because user data is very sparse due to the high volume of documents relative to the number of users. Content based approaches, on the other hand, is attractive because textual content is often very informative. In this paper we describe ...
Deep Belief Nets for Topic Modeling
2,015
http://arxiv.org/pdf/1501.04325v1
Title Deep Belief Nets Topic Modeling Summary Applying traditional collaborative filtering digital publishing challenging user data sparse due high volume document relative number user Content based approach hand attractive textual content often informative paper describe largescale content based collaborative filterin...
[0.041658274829387665, 0.0060585238970816135, -0.00080118328332901, -5.3278225095709786e-05, -0.03992142528295517, -0.006092096213251352, 0.0682005062699318, -0.01572856493294239, -0.00831314641982317, -0.007629202213138342, -0.03084690310060978, -0.0008831021841615438, -0.03961114585399628, 0.07513932883739471, -0.036...
1,740
1,740
['Rie Johnson', 'Tong Zhang']
1504.01255v3
This paper presents a new semi-supervised framework with convolutional neural networks (CNNs) for text categorization. Unlike the previous approaches that rely on word embeddings, our method learns embeddings of small text regions from unlabeled data for integration into a supervised CNN. The proposed scheme for embedd...
Semi-supervised Convolutional Neural Networks for Text Categorization via Region Embedding
2,015
http://arxiv.org/pdf/1504.01255v3
Title Semisupervised Convolutional Neural Networks Text Categorization via Region Embedding Summary paper present new semisupervised framework convolutional neural network CNNs text categorization Unlike previous approach rely word embeddings method learns embeddings small text region unlabeled data integration supervi...
[0.0377260185778141, 0.01558987982571125, 0.01853371411561966, 0.07321285456418991, -0.03153630346059799, 0.015014116652309895, 0.0369480699300766, 0.017556849867105484, 0.05739942938089371, -0.07728653401136398, -0.04226111248135567, 0.031418003141880035, -0.05130679905414581, 0.032013170421123505, -0.0039976509287953...
1,741
1,741
['Michael Bloodgood', 'John Grothendieck']
1504.06329v1
Within the natural language processing (NLP) community, active learning has been widely investigated and applied in order to alleviate the annotation bottleneck faced by developers of new NLP systems and technologies. This paper presents the first theoretical analysis of stopping active learning based on stabilizing pr...
Analysis of Stopping Active Learning based on Stabilizing Predictions
2,015
http://arxiv.org/pdf/1504.06329v1
Title Analysis Stopping Active Learning based Stabilizing Predictions Summary Within natural language processing NLP community active learning widely investigated applied order alleviate annotation bottleneck faced developer new NLP system technology paper present first theoretical analysis stopping active learning bas...
[0.051261648535728455, 0.005176495760679245, -0.018167028203606606, -0.012042935006320477, 0.0005804197280667722, 0.013916815631091595, -0.03288688138127327, 0.0040738023817539215, 0.027787398546934128, -0.07360555231571198, 0.051993485540151596, -0.008056837134063244, 0.006843406707048416, 0.06985051184892654, -0.0415...
1,742
1,742
['Shihao Ji', 'Hyokun Yun', 'Pinar Yanardag', 'Shin Matsushima', 'S. V. N. Vishwanathan']
1506.02761v4
Embedding words in a vector space has gained a lot of attention in recent years. While state-of-the-art methods provide efficient computation of word similarities via a low-dimensional matrix embedding, their motivation is often left unclear. In this paper, we argue that word embedding can be naturally viewed as a rank...
WordRank: Learning Word Embeddings via Robust Ranking
2,015
http://arxiv.org/pdf/1506.02761v4
Title WordRank Learning Word Embeddings via Robust Ranking Summary Embedding word vector space gained lot attention recent year stateoftheart method provide efficient computation word similarity via lowdimensional matrix embedding motivation often left unclear paper argue word embedding naturally viewed ranking problem...
[0.03073304519057274, 0.0170422550290823, -0.0016750956419855356, 0.06478266417980194, 0.015030897222459316, 0.013582711108028889, -0.035508036613464355, 0.005857421085238457, 0.04747035354375839, -0.05410247668623924, -0.03494616225361824, 0.006967793218791485, 0.03280114382505417, -0.010566591285169125, 0.05743702128...
1,743
1,743
['Jacob Andreas', 'Maxim Rabinovich', 'Dan Klein', 'Michael I. Jordan']
1506.04147v2
Calculation of the log-normalizer is a major computational obstacle in applications of log-linear models with large output spaces. The problem of fast normalizer computation has therefore attracted significant attention in the theoretical and applied machine learning literature. In this paper, we analyze a recently pro...
On the accuracy of self-normalized log-linear models
2,015
http://arxiv.org/pdf/1506.04147v2
Title accuracy selfnormalized loglinear model Summary Calculation lognormalizer major computational obstacle application loglinear model large output space problem fast normalizer computation therefore attracted significant attention theoretical applied machine learning literature paper analyze recently proposed techni...
[-0.050399210304021835, 0.014546597376465797, -0.006212917156517506, -0.015343377366662025, 0.0012597260065376759, -0.025381935760378838, 0.050457607954740524, -0.002719401614740491, -0.03146839886903763, 0.0037624090909957886, -0.015648124739527702, 0.025613514706492424, -0.046493083238601685, 0.059850044548511505, 0....
1,744
1,744
['Jiatao Gu', 'Victor O. K. Li']
1506.07477v1
Replicated Softmax model, a well-known undirected topic model, is powerful in extracting semantic representations of documents. Traditional learning strategies such as Contrastive Divergence are very inefficient. This paper provides a novel estimator to speed up the learning based on Noise Contrastive Estimate, extende...
Efficient Learning for Undirected Topic Models
2,015
http://arxiv.org/pdf/1506.07477v1
Title Efficient Learning Undirected Topic Models Summary Replicated Softmax model wellknown undirected topic model powerful extracting semantic representation document Traditional learning strategy Contrastive Divergence inefficient paper provides novel estimator speed learning based Noise Contrastive Estimate extended...
[0.03385142236948013, -0.03591778501868248, 0.011517507024109364, 0.020374715328216553, -0.018401702865958214, 0.02540573664009571, 0.012628433294594288, 0.03921886160969734, -0.06747400760650635, -0.06107277423143387, -0.03184773027896881, 0.02360793948173523, -0.017412235960364342, 0.014786489307880402, -0.0076919537...
1,745
1,745
['Alex Auvolat', 'Sarath Chandar', 'Pascal Vincent', 'Hugo Larochelle', 'Yoshua Bengio']
1507.05910v3
Efficient Maximum Inner Product Search (MIPS) is an important task that has a wide applicability in recommendation systems and classification with a large number of classes. Solutions based on locality-sensitive hashing (LSH) as well as tree-based solutions have been investigated in the recent literature, to perform ap...
Clustering is Efficient for Approximate Maximum Inner Product Search
2,015
http://arxiv.org/pdf/1507.05910v3
Title Clustering Efficient Approximate Maximum Inner Product Search Summary Efficient Maximum Inner Product Search MIPS important task wide applicability recommendation system classification large number class Solutions based localitysensitive hashing LSH well treebased solution investigated recent literature perform a...
[-0.019958825781941414, -0.04636021703481674, -0.005257997661828995, 0.026999106630682945, 0.026350820437073708, -0.01876971311867237, 0.025029445067048073, 0.06791116297245026, 0.03354522958397865, -0.03609990328550339, -0.052353084087371826, 0.010323484428226948, -0.030700473114848137, -0.001563530066050589, -0.02752...
1,746
1,746
['Shuangyin Li', 'Jiefei Li', 'Guan Huang', 'Ruiyang Tan', 'Rong Pan']
1507.08396v1
To date, there have been massive Semi-Structured Documents (SSDs) during the evolution of the Internet. These SSDs contain both unstructured features (e.g., plain text) and metadata (e.g., tags). Most previous works focused on modeling the unstructured text, and recently, some other methods have been proposed to model ...
Tag-Weighted Topic Model For Large-scale Semi-Structured Documents
2,015
http://arxiv.org/pdf/1507.08396v1
Title TagWeighted Topic Model Largescale SemiStructured Documents Summary date massive SemiStructured Documents SSDs evolution Internet SSDs contain unstructured feature eg plain text metadata eg tag previous work focused modeling unstructured text recently method proposed model unstructured text specific tag build gen...
[0.044244103133678436, -0.010632669553160667, -0.01801956258714199, -0.002537464490160346, -0.021081242710351944, 0.017650030553340912, -0.006492476910352707, 0.012165571562945843, 0.009466796182096004, -0.06178731098771095, -0.010671790689229965, -0.0006670161383226514, -0.030175384134054184, 0.05782283470034599, -0.0...
1,747
1,747
['Peixian Chen', 'Nevin L. Zhang', 'Leonard K. M. Poon', 'Zhourong Chen']
1508.00973v1
Hierarchical latent tree analysis (HLTA) is recently proposed as a new method for topic detection. It differs fundamentally from the LDA-based methods in terms of topic definition, topic-document relationship, and learning method. It has been shown to discover significantly more coherent topics and better topic hierarc...
Progressive EM for Latent Tree Models and Hierarchical Topic Detection
2,015
http://arxiv.org/pdf/1508.00973v1
Title Progressive EM Latent Tree Models Hierarchical Topic Detection Summary Hierarchical latent tree analysis HLTA recently proposed new method topic detection differs fundamentally LDAbased method term topic definition topicdocument relationship learning method shown discover significantly coherent topic better topic...
[0.03521453216671944, -0.0005978510016575456, -0.01671707071363926, 0.05056741088628769, -0.00897853635251522, 0.032421767711639404, -0.05680364370346069, 0.014174436219036579, -0.06642012298107147, -0.02821076288819313, 0.01717168092727661, 0.010218137875199318, -0.02928110770881176, 0.023723820224404335, -0.036467637...
1,748
1,748
['Ralph Fehrer', 'Stefan Feuerriegel']
1508.01993v1
Decision analytics commonly focuses on the text mining of financial news sources in order to provide managerial decision support and to predict stock market movements. Existing predictive frameworks almost exclusively apply traditional machine learning methods, whereas recent research indicates that traditional machine...
Improving Decision Analytics with Deep Learning: The Case of Financial Disclosures
2,015
http://arxiv.org/pdf/1508.01993v1
Title Improving Decision Analytics Deep Learning Case Financial Disclosures Summary Decision analytics commonly focus text mining financial news source order provide managerial decision support predict stock market movement Existing predictive framework almost exclusively apply traditional machine learning method where...
[-0.01072494313120842, 0.05536654219031334, -0.04572255536913872, -0.023944061249494553, -0.00735219893977046, -0.015500505454838276, -0.0013168975710868835, 0.030248241499066353, -0.0060745603404939175, -0.02557271160185337, 0.019153334200382233, 0.031049814075231552, -0.011908790096640587, 0.10163259506225586, 0.0077...
1,749
1,749
['Shaohua Li', 'Jun Zhu', 'Chunyan Miao']
1508.03826v1
Most existing word embedding methods can be categorized into Neural Embedding Models and Matrix Factorization (MF)-based methods. However some models are opaque to probabilistic interpretation, and MF-based methods, typically solved using Singular Value Decomposition (SVD), may incur loss of corpus information. In addi...
A Generative Word Embedding Model and its Low Rank Positive Semidefinite Solution
2,015
http://arxiv.org/pdf/1508.03826v1
Title Generative Word Embedding Model Low Rank Positive Semidefinite Solution Summary existing word embedding method categorized Neural Embedding Models Matrix Factorization MFbased method However model opaque probabilistic interpretation MFbased method typically solved using Singular Value Decomposition SVD may incur ...
[0.019701749086380005, 0.04174591600894928, 0.0020315188448876143, 0.05907386168837547, 0.009612975642085075, 0.0043312180787324905, -0.027331162244081497, -0.008907875046133995, -0.008909355849027634, -0.04644520580768585, 0.01359839178621769, 0.002051933901384473, 0.011301317252218723, 0.03821622580289841, 0.02160174...
1,750
1,750
['Weicong Ding', 'Prakash Ishwar', 'Venkatesh Saligrama']
1508.05565v2
We develop necessary and sufficient conditions and a novel provably consistent and efficient algorithm for discovering topics (latent factors) from observations (documents) that are realized from a probabilistic mixture of shared latent factors that have certain properties. Our focus is on the class of topic models in ...
Necessary and Sufficient Conditions and a Provably Efficient Algorithm for Separable Topic Discovery
2,015
http://arxiv.org/pdf/1508.05565v2
Title Necessary Sufficient Conditions Provably Efficient Algorithm Separable Topic Discovery Summary develop necessary sufficient condition novel provably consistent efficient algorithm discovering topic latent factor observation document realized probabilistic mixture shared latent factor certain property focus class ...
[0.017968598753213882, 0.02832053229212761, -0.01064675860106945, 0.06470052897930145, -0.04353935271501541, -0.005210284609347582, -0.0013501578941941261, 0.0036546350456774235, -0.048898257315158844, -0.018110215663909912, 0.0025604492984712124, 0.037263695150613785, 0.009986727498471737, 0.07320576906204224, -0.0430...
1,751
1,751
['Simon Šuster', 'Gertjan van Noord', 'Ivan Titov']
1508.07709v2
Word representations induced from models with discrete latent variables (e.g.\ HMMs) have been shown to be beneficial in many NLP applications. In this work, we exploit labeled syntactic dependency trees and formalize the induction problem as unsupervised learning of tree-structured hidden Markov models. Syntactic func...
Word Representations, Tree Models and Syntactic Functions
2,015
http://arxiv.org/pdf/1508.07709v2
Title Word Representations Tree Models Syntactic Functions Summary Word representation induced model discrete latent variable eg HMMs shown beneficial many NLP application work exploit labeled syntactic dependency tree formalize induction problem unsupervised learning treestructured hidden Markov model Syntactic functi...
[0.041282642632722855, 0.000679301971103996, 0.0035394069273024797, 0.06432737410068512, -0.03604457899928093, 0.01592227630317211, -0.01988198049366474, 0.022751769050955772, 0.003970311023294926, -0.1323225498199463, 0.01592140644788742, 0.010402299463748932, 0.03395913541316986, 0.02736915461719036, -0.0043000602163...
1,752
1,752
['Tatsunori B. Hashimoto', 'David Alvarez-Melis', 'Tommi S. Jaakkola']
1509.05808v1
Continuous vector representations of words and objects appear to carry surprisingly rich semantic content. In this paper, we advance both the conceptual and theoretical understanding of word embeddings in three ways. First, we ground embeddings in semantic spaces studied in cognitive-psychometric literature and introdu...
Word, graph and manifold embedding from Markov processes
2,015
http://arxiv.org/pdf/1509.05808v1
Title Word graph manifold embedding Markov process Summary Continuous vector representation word object appear carry surprisingly rich semantic content paper advance conceptual theoretical understanding word embeddings three way First ground embeddings semantic space studied cognitivepsychometric literature introduce n...
[0.005375457927584648, -0.02207511104643345, -0.0016465680673718452, 0.0387234091758728, -0.03534363955259323, 0.019127370789647102, -0.04923190176486969, 0.012723374180495739, 0.017417293041944504, -0.05896516889333725, 0.024966957047581673, 0.017224501818418503, 0.035561397671699524, -0.005777963437139988, 0.02925400...
1,753
1,753
['Kai-Wei Chang', 'Shyam Upadhyay', 'Ming-Wei Chang', 'Vivek Srikumar', 'Dan Roth']
1509.07179v1
IllinoisSL is a Java library for learning structured prediction models. It supports structured Support Vector Machines and structured Perceptron. The library consists of a core learning module and several applications, which can be executed from command-lines. Documentation is provided to guide users. In Comparison to ...
IllinoisSL: A JAVA Library for Structured Prediction
2,015
http://arxiv.org/pdf/1509.07179v1
Title IllinoisSL JAVA Library Structured Prediction Summary IllinoisSL Java library learning structured prediction model support structured Support Vector Machines structured Perceptron library consists core learning module several application executed commandlines Documentation provided guide user Comparison structure...
[0.0054570818319916725, 0.01697663962841034, -0.025806888937950134, 0.0032336486037820578, -0.01236091461032629, -0.0027404797729104757, 0.04882329702377319, 0.047030579298734665, 0.03444981947541237, -0.020768150687217712, 0.0451328419148922, 0.021536987274885178, 0.032788459211587906, 0.05362945795059204, -0.00991686...
1,754
1,754
['Stephanie L. Hyland', 'Theofanis Karaletsos', 'Gunnar Rätsch']
1510.00259v2
Neural language models are a powerful tool to embed words into semantic vector spaces. However, learning such models generally relies on the availability of abundant and diverse training examples. In highly specialised domains this requirement may not be met due to difficulties in obtaining a large corpus, or the limit...
A Generative Model of Words and Relationships from Multiple Sources
2,015
http://arxiv.org/pdf/1510.00259v2
Title Generative Model Words Relationships Multiple Sources Summary Neural language model powerful tool embed word semantic vector space However learning model generally relies availability abundant diverse training example highly specialised domain requirement may met due difficulty obtaining large corpus limited rang...
[0.04936066269874573, 0.059426575899124146, -0.00502361822873354, 0.04834246262907982, -0.01448764931410551, -0.006121241021901369, 0.003559398464858532, -0.001421733177267015, -0.03364616632461548, -0.04608825594186783, -0.008568217046558857, -0.005939324386417866, 0.01180427148938179, 0.05401214584708214, 0.036159686...
1,755
1,755
['Osama Khalifa', 'David Wolfe Corne', 'Mike Chantler']
1510.06646v2
Hyper-parameters play a major role in the learning and inference process of latent Dirichlet allocation (LDA). In order to begin the LDA latent variables learning process, these hyper-parameters values need to be pre-determined. We propose an extension for LDA that we call 'Latent Dirichlet allocation Gibbs Newton' (LD...
A 'Gibbs-Newton' Technique for Enhanced Inference of Multivariate Polya Parameters and Topic Models
2,015
http://arxiv.org/pdf/1510.06646v2
Title GibbsNewton Technique Enhanced Inference Multivariate Polya Parameters Topic Models Summary Hyperparameters play major role learning inference process latent Dirichlet allocation LDA order begin LDA latent variable learning process hyperparameters value need predetermined propose extension LDA call Latent Dirichl...
[0.04304962605237961, 0.03057274781167507, -0.011775540187954903, 0.003363706637173891, -0.01754011958837509, 0.020197579637169838, 0.027440108358860016, -0.006536873988807201, -0.03836359828710556, -0.07477635890245438, 0.023660559207201004, 0.006086000241339207, 0.010895522311329842, 0.009973455220460892, -0.01184165...
1,756
1,756
['Abhinav Maurya']
1511.00352v3
Many methods have been proposed for detecting emerging events in text streams using topic modeling. However, these methods have shortcomings that make them unsuitable for rapid detection of locally emerging events on massive text streams. We describe Spatially Compact Semantic Scan (SCSS) that has been developed specif...
Spatial Semantic Scan: Jointly Detecting Subtle Events and their Spatial Footprint
2,015
http://arxiv.org/pdf/1511.00352v3
Title Spatial Semantic Scan Jointly Detecting Subtle Events Spatial Footprint Summary Many method proposed detecting emerging event text stream using topic modeling However method shortcoming make unsuitable rapid detection locally emerging event massive text stream describe Spatially Compact Semantic Scan SCSS develop...
[0.02946445345878601, -0.016615962609648705, -0.02140660397708416, 0.01176355592906475, -0.048396382480859756, -0.016953468322753906, -0.0052398452535271645, 0.013773884624242783, -0.031840093433856964, -0.05497649312019348, 0.06431789696216583, 0.024369925260543823, -0.002895390847697854, 0.07184825092554092, -0.03930...
1,757
1,757
['Yangfeng Ji', 'Trevor Cohn', 'Lingpeng Kong', 'Chris Dyer', 'Jacob Eisenstein']
1511.03962v4
Text documents are structured on multiple levels of detail: individual words are related by syntax, but larger units of text are related by discourse structure. Existing language models generally fail to account for discourse structure, but it is crucial if we are to have language models that reward coherence and gener...
Document Context Language Models
2,015
http://arxiv.org/pdf/1511.03962v4
Title Document Context Language Models Summary Text document structured multiple level detail individual word related syntax larger unit text related discourse structure Existing language model generally fail account discourse structure crucial language model reward coherence generate coherent text present empirically ...
[0.021096419543027878, 0.019633939489722252, 0.013651952147483826, 0.032060664147138596, -0.055084336549043655, -0.006428249645978212, 0.03714514523744583, -0.024491848424077034, -0.04989150911569595, -0.07053882628679276, -0.009069678373634815, -0.0067405253648757935, 0.050534676760435104, 0.047179169952869415, 0.0029...
1,758
1,758
['Arvind Neelakantan', 'Quoc V. Le', 'Ilya Sutskever']
1511.04834v3
Deep neural networks have achieved impressive supervised classification performance in many tasks including image recognition, speech recognition, and sequence to sequence learning. However, this success has not been translated to applications like question answering that may involve complex arithmetic and logic reason...
Neural Programmer: Inducing Latent Programs with Gradient Descent
2,015
http://arxiv.org/pdf/1511.04834v3
Title Neural Programmer Inducing Latent Programs Gradient Descent Summary Deep neural network achieved impressive supervised classification performance many task including image recognition speech recognition sequence sequence learning However success translated application like question answering may involve complex a...
[0.009041440673172474, 0.08065886795520782, -0.024964742362499237, 0.02497670240700245, 0.004030184354633093, 0.012189928442239761, 0.014649591408669949, -0.05598830059170723, -0.05914132669568062, -0.013189796358346939, -0.0029175058007240295, 0.06548763811588287, 0.006712948903441429, 0.05989823862910271, 0.043166257...
1,759
1,759
['Eric Nalisnick', 'Sachin Ravi']
1511.05392v3
We describe a method for learning word embeddings with data-dependent dimensionality. Our Stochastic Dimensionality Skip-Gram (SD-SG) and Stochastic Dimensionality Continuous Bag-of-Words (SD-CBOW) are nonparametric analogs of Mikolov et al.'s (2013) well-known 'word2vec' models. Vector dimensionality is made dynamic b...
Learning the Dimensionality of Word Embeddings
2,015
http://arxiv.org/pdf/1511.05392v3
Title Learning Dimensionality Word Embeddings Summary describe method learning word embeddings datadependent dimensionality Stochastic Dimensionality SkipGram SDSG Stochastic Dimensionality Continuous BagofWords SDCBOW nonparametric analog Mikolov et al 2013 wellknown word2vec model Vector dimensionality made dynamic e...
[0.004331451375037432, 0.004026761278510094, -0.00014245757483877242, 0.0659000501036644, -0.00292156427167356, 0.011031015776097775, -0.005850176326930523, 0.012247947975993156, -0.021671665832400322, -0.027764195576310158, 0.041022635996341705, -0.04336189851164818, 0.025085369125008583, 0.05910046026110649, 0.040243...
1,760
1,760
['Yishu Miao', 'Lei Yu', 'Phil Blunsom']
1511.06038v4
Recent advances in neural variational inference have spawned a renaissance in deep latent variable models. In this paper we introduce a generic variational inference framework for generative and conditional models of text. While traditional variational methods derive an analytic approximation for the intractable distri...
Neural Variational Inference for Text Processing
2,015
http://arxiv.org/pdf/1511.06038v4
Title Neural Variational Inference Text Processing Summary Recent advance neural variational inference spawned renaissance deep latent variable model paper introduce generic variational inference framework generative conditional model text traditional variational method derive analytic approximation intractable distrib...
[0.05716373398900032, 0.09118175506591797, -0.020867841318249702, 0.026435362175107002, -0.028947971761226654, 0.007027584593743086, -0.0011308362009003758, -0.01164302509278059, -0.01497575081884861, 0.006230594124644995, 0.034897226840257645, -0.017986750230193138, -0.022486459463834763, 0.05332551151514053, 0.043549...
1,761
1,761
['Minh-Thang Luong', 'Quoc V. Le', 'Ilya Sutskever', 'Oriol Vinyals', 'Lukasz Kaiser']
1511.06114v4
Sequence to sequence learning has recently emerged as a new paradigm in supervised learning. To date, most of its applications focused on only one task and not much work explored this framework for multiple tasks. This paper examines three multi-task learning (MTL) settings for sequence to sequence models: (a) the onet...
Multi-task Sequence to Sequence Learning
2,015
http://arxiv.org/pdf/1511.06114v4
Title Multitask Sequence Sequence Learning Summary Sequence sequence learning recently emerged new paradigm supervised learning date application focused one task much work explored framework multiple task paper examines three multitask learning MTL setting sequence sequence model onetomany setting encoder shared severa...
[0.023004015907645226, 0.014320282265543938, -0.01768209971487522, 0.030446134507656097, -0.029457369819283485, 0.03642934188246727, 0.031160037964582443, -0.026653746142983437, -0.018277036026120186, -0.048788998275995255, -0.012121399864554405, -0.018034644424915314, -0.011738212779164314, 0.05399543419480324, 0.0088...
1,762
1,762
['Oriol Vinyals', 'Samy Bengio', 'Manjunath Kudlur']
1511.06391v4
Sequences have become first class citizens in supervised learning thanks to the resurgence of recurrent neural networks. Many complex tasks that require mapping from or to a sequence of observations can now be formulated with the sequence-to-sequence (seq2seq) framework which employs the chain rule to efficiently repre...
Order Matters: Sequence to sequence for sets
2,015
http://arxiv.org/pdf/1511.06391v4
Title Order Matters Sequence sequence set Summary Sequences become first class citizen supervised learning thanks resurgence recurrent neural network Many complex task require mapping sequence observation formulated sequencetosequence seq2seq framework employ chain rule efficiently represent joint probability sequence ...
[0.02856549806892872, 0.05684865266084671, -0.013269457034766674, 0.03685982897877693, -0.008908270858228207, 0.015288462862372398, -0.013774442486464977, -0.02104431390762329, -0.011418582871556282, -0.03381778672337532, 0.044366247951984406, -0.031216273084282875, 0.04467383027076721, 0.0632706955075264, 0.0201888289...
1,763
1,763
['Sanjeev Arora', 'Yuanzhi Li', 'Yingyu Liang', 'Tengyu Ma', 'Andrej Risteski']
1601.03764v2
Word embeddings are ubiquitous in NLP and information retrieval, but it's unclear what they represent when the word is polysemous, i.e., has multiple senses. Here it is shown that multiple word senses reside in linear superposition within the word embedding and can be recovered by simple sparse coding. The success of...
Linear Algebraic Structure of Word Senses, with Applications to Polysemy
2,016
http://arxiv.org/pdf/1601.03764v2
Title Linear Algebraic Structure Word Senses Applications Polysemy Summary Word embeddings ubiquitous NLP information retrieval unclear represent word polysemous ie multiple sens shown multiple word sens reside linear superposition within word embedding recovered simple sparse coding success method applies several embe...
[0.02782212384045124, 0.019674578681588173, -0.020404214039444923, 0.07692492753267288, -0.012704483233392239, 0.009400893934071064, -0.01284840703010559, 0.02990558370947838, 0.026957964524626732, -0.03731299936771393, 0.02244720421731472, 0.0002192384999943897, 0.041244279593229294, 0.03909743204712868, 0.00973379146...
1,764
1,764
['Pranay Dighe', 'Gil Luyet', 'Afsaneh Asaei', 'Herve Bourlard']
1601.05936v1
We propose to model the acoustic space of deep neural network (DNN) class-conditional posterior probabilities as a union of low-dimensional subspaces. To that end, the training posteriors are used for dictionary learning and sparse coding. Sparse representation of the test posteriors using this dictionary enables proje...
Exploiting Low-dimensional Structures to Enhance DNN Based Acoustic Modeling in Speech Recognition
2,016
http://arxiv.org/pdf/1601.05936v1
Title Exploiting Lowdimensional Structures Enhance DNN Based Acoustic Modeling Speech Recognition Summary propose model acoustic space deep neural network DNN classconditional posterior probability union lowdimensional subspace end training posterior used dictionary learning sparse coding Sparse representation test pos...
[-0.025205491110682487, 0.05401929095387459, 0.008178629912436008, 0.03717883676290512, 0.0028122535441070795, -0.0014283980708569288, 0.04569118842482567, -0.01851784624159336, -0.03298197314143181, -0.018088048323988914, -0.04133128374814987, -0.015296448953449726, 0.06723323464393616, 0.028978973627090454, 0.0303332...
1,765
1,765
['André F. T. Martins', 'Ramón Fernandez Astudillo']
1602.02068v2
We propose sparsemax, a new activation function similar to the traditional softmax, but able to output sparse probabilities. After deriving its properties, we show how its Jacobian can be efficiently computed, enabling its use in a network trained with backpropagation. Then, we propose a new smooth and convex loss func...
From Softmax to Sparsemax: A Sparse Model of Attention and Multi-Label Classification
2,016
http://arxiv.org/pdf/1602.02068v2
Title Softmax Sparsemax Sparse Model Attention MultiLabel Classification Summary propose sparsemax new activation function similar traditional softmax able output sparse probability deriving property show Jacobian efficiently computed enabling use network trained backpropagation propose new smooth convex loss function ...
[0.043849878013134, -0.0018448815681040287, 0.0061729722656309605, 0.06235148385167122, 0.030350953340530396, -0.010795783251523972, 0.020317701622843742, -0.01468266174197197, -0.012204651720821857, -0.04486563801765442, -0.04551755264401436, -0.009337262250483036, 0.011911483481526375, 0.026033539324998856, 0.0499399...
1,766
1,766
['Rie Johnson', 'Tong Zhang']
1602.02373v2
One-hot CNN (convolutional neural network) has been shown to be effective for text categorization (Johnson & Zhang, 2015). We view it as a special case of a general framework which jointly trains a linear model with a non-linear feature generator consisting of `text region embedding + pooling'. Under this framework, we...
Supervised and Semi-Supervised Text Categorization using LSTM for Region Embeddings
2,016
http://arxiv.org/pdf/1602.02373v2
Title Supervised SemiSupervised Text Categorization using LSTM Region Embeddings Summary Onehot CNN convolutional neural network shown effective text categorization Johnson Zhang 2015 view special case general framework jointly train linear model nonlinear feature generator consisting text region embedding pooling fram...
[0.02528824470937252, 0.0048169041983783245, 0.0048355585895478725, 0.06478418409824371, -0.018797703087329865, 0.00879511795938015, 0.05919131264090538, 0.023448346182703972, 0.06189616397023201, -0.054228901863098145, -0.044883180409669876, 0.005527803208678961, -0.039715301245450974, 0.03627880662679672, 0.006763240...
1,767
1,767
['Bo Tang', 'Steven Kay', 'Haibo He']
1602.02850v1
Automated feature selection is important for text categorization to reduce the feature size and to speed up the learning process of classifiers. In this paper, we present a novel and efficient feature selection framework based on the Information Theory, which aims to rank the features with their discriminative capacity...
Toward Optimal Feature Selection in Naive Bayes for Text Categorization
2,016
http://arxiv.org/pdf/1602.02850v1
Title Toward Optimal Feature Selection Naive Bayes Text Categorization Summary Automated feature selection important text categorization reduce feature size speed learning process classifier paper present novel efficient feature selection framework based Information Theory aim rank feature discriminative capacity class...
[0.0341770239174366, -0.015858760103583336, -0.012375161983072758, 0.007200086489319801, -0.03821850195527077, 0.04265877231955528, 0.049862343817949295, 0.06063368171453476, 0.03556988760828972, -0.08319222927093506, 0.006110100075602531, 0.011010958813130856, 0.01946283131837845, 0.004525516182184219, -0.046380080282...
1,768
1,768
['Yong Ren', 'Yining Wang', 'Jun Zhu']
1602.06025v1
Supervised topic models simultaneously model the latent topic structure of large collections of documents and a response variable associated with each document. Existing inference methods are based on variational approximation or Monte Carlo sampling, which often suffers from the local minimum defect. Spectral methods ...
Spectral Learning for Supervised Topic Models
2,016
http://arxiv.org/pdf/1602.06025v1
Title Spectral Learning Supervised Topic Models Summary Supervised topic model simultaneously model latent topic structure large collection document response variable associated document Existing inference method based variational approximation Monte Carlo sampling often suffers local minimum defect Spectral method app...
[0.0425785630941391, 0.020060230046510696, -0.021234748885035515, 0.0221326295286417, -0.02517617680132389, -0.02801104448735714, -0.02258664183318615, -0.0005783329834230244, -0.044709980487823486, -0.030400056391954422, -0.013388658873736858, 0.023481015115976334, 0.003824490588158369, 0.03611167520284653, 0.00286192...
1,769
1,769
['Saurav Ghosh', 'Prithwish Chakraborty', 'Emily Cohn', 'John S. Brownstein', 'Naren Ramakrishnan']
1603.00106v2
Traditional disease surveillance can be augmented with a wide variety of real-time sources such as, news and social media. However, these sources are in general unstructured and, construction of surveillance tools such as taxonomical correlations and trace mapping involves considerable human supervision. In this paper,...
Characterizing Diseases from Unstructured Text: A Vocabulary Driven Word2vec Approach
2,016
http://arxiv.org/pdf/1603.00106v2
Title Characterizing Diseases Unstructured Text Vocabulary Driven Word2vec Approach Summary Traditional disease surveillance augmented wide variety realtime source news social medium However source general unstructured construction surveillance tool taxonomical correlation trace mapping involves considerable human supe...
[0.03265061974525452, 0.050908081233501434, 0.006352902390062809, -0.034371525049209595, 0.04244180768728256, 0.03581146150827408, -0.03879131004214287, 0.03228770196437836, 0.0005203719483688474, -0.03878666087985039, 0.04945071041584015, -0.01110029499977827, 0.00811578705906868, 0.04277942702174187, 0.00204042065888...
1,770
1,770
['Xuezhe Ma', 'Eduard Hovy']
1603.01354v5
State-of-the-art sequence labeling systems traditionally require large amounts of task-specific knowledge in the form of hand-crafted features and data pre-processing. In this paper, we introduce a novel neutral network architecture that benefits from both word- and character-level representations automatically, by usi...
End-to-end Sequence Labeling via Bi-directional LSTM-CNNs-CRF
2,016
http://arxiv.org/pdf/1603.01354v5
Title Endtoend Sequence Labeling via Bidirectional LSTMCNNsCRF Summary Stateoftheart sequence labeling system traditionally require large amount taskspecific knowledge form handcrafted feature data preprocessing paper introduce novel neutral network architecture benefit word characterlevel representation automatically ...
[0.03140919283032417, 0.036552418023347855, 0.03543682023882866, 0.025777094066143036, -0.03740682825446129, 0.03663747012615204, -0.0040022521279752254, -6.026535629644059e-06, -0.0038274643011391163, -0.05678362399339676, -0.01461838185787201, -0.04551257565617561, -0.0007777315331622958, 0.058694932609796524, 0.0112...
1,771
1,771
['Mike Kestemont', 'Jeroen De Gussem']
1603.01597v2
In this paper we consider two sequence tagging tasks for medieval Latin: part-of-speech tagging and lemmatization. These are both basic, yet foundational preprocessing steps in applications such as text re-use detection. Nevertheless, they are generally complicated by the considerable orthographic variation which is ty...
Integrated Sequence Tagging for Medieval Latin Using Deep Representation Learning
2,016
http://arxiv.org/pdf/1603.01597v2
Title Integrated Sequence Tagging Medieval Latin Using Deep Representation Learning Summary paper consider two sequence tagging task medieval Latin partofspeech tagging lemmatization basic yet foundational preprocessing step application text reuse detection Nevertheless generally complicated considerable orthographic v...
[-0.01326470822095871, 0.05052555724978447, -0.028928382322192192, 0.0713241845369339, -0.04272061213850975, 0.023121684789657593, 0.017576362937688828, 0.019176635891199112, 0.019660213962197304, -0.07454778999090195, 0.03953102231025696, -0.021148161962628365, 0.05133505538105965, -0.016845576465129852, 0.03358857333...
1,772
1,772
['Simon Šuster', 'Ivan Titov', 'Gertjan van Noord']
1603.09128v1
We present an approach to learning multi-sense word embeddings relying both on monolingual and bilingual information. Our model consists of an encoder, which uses monolingual and bilingual context (i.e. a parallel sentence) to choose a sense for a given word, and a decoder which predicts context words based on the chos...
Bilingual Learning of Multi-sense Embeddings with Discrete Autoencoders
2,016
http://arxiv.org/pdf/1603.09128v1
Title Bilingual Learning Multisense Embeddings Discrete Autoencoders Summary present approach learning multisense word embeddings relying monolingual bilingual information model consists encoder us monolingual bilingual context ie parallel sentence choose sense given word decoder predicts context word based chosen sens...
[-0.030703021213412285, 0.02101733908057213, -0.02560856007039547, 0.07318764925003052, -0.026089604943990707, 0.041040387004613876, 0.04532865434885025, -0.05407891422510147, 0.0009108321974053979, -0.05324386805295944, -0.03569773957133293, -0.03188089281320572, 0.011515633203089237, -0.019597431644797325, 0.01825074...
1,773
1,773
['Bin Wang', 'Zhijian Ou', 'Yong He', 'Akinori Kawamura']
1603.09170v5
The dominant language models (LMs) such as n-gram and neural network (NN) models represent sentence probabilities in terms of conditionals. In contrast, a new trans-dimensional random field (TRF) LM has been recently introduced to show superior performances, where the whole sentence is modeled as a random field. In thi...
Model Interpolation with Trans-dimensional Random Field Language Models for Speech Recognition
2,016
http://arxiv.org/pdf/1603.09170v5
Title Model Interpolation Transdimensional Random Field Language Models Speech Recognition Summary dominant language model LMs ngram neural network NN model represent sentence probability term conditionals contrast new transdimensional random field TRF LM recently introduced show superior performance whole sentence mod...
[0.03344308212399483, 0.04040622338652611, 0.02594810351729393, 0.06121968477964401, -0.05249654874205589, -0.015675010159611702, -0.003350897692143917, 0.009241187945008278, -0.026914099231362343, -0.06283418834209442, -0.013268359936773777, -0.04543420299887657, 0.01214163564145565, 0.02181263454258442, 0.00875791627...
1,774
1,774
['Ricardo Pio Monti', 'Romy Lorenz', 'Robert Leech', 'Christoforos Anagnostopoulos', 'Giovanni Montana']
1605.00223v1
Large-scale automated meta-analysis of neuroimaging data has recently established itself as an important tool in advancing our understanding of human brain function. This research has been pioneered by NeuroSynth, a database collecting both brain activation coordinates and associated text across a large cohort of neuro...
Text-mining the NeuroSynth corpus using Deep Boltzmann Machines
2,016
http://arxiv.org/pdf/1605.00223v1
Title Textmining NeuroSynth corpus using Deep Boltzmann Machines Summary Largescale automated metaanalysis neuroimaging data recently established important tool advancing understanding human brain function research pioneered NeuroSynth database collecting brain activation coordinate associated text across large cohort ...
[0.019228285178542137, 0.01063005905598402, -0.021334486082196236, 0.0010391651885583997, 0.005581077188253403, 0.052387211471796036, 0.07854806631803513, 0.01290129404515028, -0.02477327175438404, 0.004262220114469528, -0.008952469564974308, -0.009977017529308796, 0.010991971008479595, 0.07068359106779099, 0.031701575...
1,775
1,775
['Kyunghyun Cho']
1605.03835v1
Recent advances in conditional recurrent language modelling have mainly focused on network architectures (e.g., attention mechanism), learning algorithms (e.g., scheduled sampling and sequence-level training) and novel applications (e.g., image/video description generation, speech recognition, etc.) On the other hand, ...
Noisy Parallel Approximate Decoding for Conditional Recurrent Language Model
2,016
http://arxiv.org/pdf/1605.03835v1
Title Noisy Parallel Approximate Decoding Conditional Recurrent Language Model Summary Recent advance conditional recurrent language modelling mainly focused network architecture eg attention mechanism learning algorithm eg scheduled sampling sequencelevel training novel application eg imagevideo description generation...
[0.022052796557545662, 0.04077897593379021, 0.011505876667797565, 0.08497445285320282, -0.018931305035948753, 0.02187930978834629, -0.011630714870989323, -0.024418912827968597, -0.02059754729270935, -0.01032083947211504, -0.017650509253144264, -0.04753543809056282, 0.009193453006446362, 0.014381315559148788, -0.0009750...
1,776
1,776
['Peixian Chen', 'Nevin L. Zhang', 'Tengfei Liu', 'Leonard K. M. Poon', 'Zhourong Chen', 'Farhan Khawar']
1605.06650v2
We present a novel method for hierarchical topic detection where topics are obtained by clustering documents in multiple ways. Specifically, we model document collections using a class of graphical models called hierarchical latent tree models (HLTMs). The variables at the bottom level of an HLTM are observed binary va...
Latent Tree Models for Hierarchical Topic Detection
2,016
http://arxiv.org/pdf/1605.06650v2
Title Latent Tree Models Hierarchical Topic Detection Summary present novel method hierarchical topic detection topic obtained clustering document multiple way Specifically model document collection using class graphical model called hierarchical latent tree model HLTMs variable bottom level HLTM observed binary variab...
[0.043696627020835876, -0.018787406384944916, -0.027211656793951988, 0.04346488416194916, -0.0422467440366745, 0.012872815132141113, -0.03228652849793434, 0.03967881575226784, -0.03319607675075531, -0.04977014288306236, 0.012331594713032246, 0.021323254331946373, -0.007156564388424158, 0.051743194460868835, -0.03271131...
1,777
1,777
['Artem Sokolov', 'Julia Kreutzer', 'Christopher Lo', 'Stefan Riezler']
1606.00739v2
Stochastic structured prediction under bandit feedback follows a learning protocol where on each of a sequence of iterations, the learner receives an input, predicts an output structure, and receives partial feedback in form of a task loss evaluation of the predicted structure. We present applications of this learning ...
Stochastic Structured Prediction under Bandit Feedback
2,016
http://arxiv.org/pdf/1606.00739v2
Title Stochastic Structured Prediction Bandit Feedback Summary Stochastic structured prediction bandit feedback follows learning protocol sequence iteration learner receives input predicts output structure receives partial feedback form task loss evaluation predicted structure present application learning scenario conv...
[0.060265157371759415, 0.03925081714987755, 0.026947712525725365, 0.030523257330060005, -0.008683743886649609, 0.0005289496039040387, -0.026373490691184998, 0.0390591137111187, -0.0032415043096989393, -0.0354313850402832, -0.029853442683815956, -0.033773571252822876, 0.00322741293348372, 0.07418493926525116, 0.02379938...
1,778
1,778
['Viktoriya Krakovna', 'Finale Doshi-Velez']
1606.05320v2
As deep neural networks continue to revolutionize various application domains, there is increasing interest in making these powerful models more understandable and interpretable, and narrowing down the causes of good and bad predictions. We focus on recurrent neural networks (RNNs), state of the art models in speech re...
Increasing the Interpretability of Recurrent Neural Networks Using Hidden Markov Models
2,016
http://arxiv.org/pdf/1606.05320v2
Title Increasing Interpretability Recurrent Neural Networks Using Hidden Markov Models Summary deep neural network continue revolutionize various application domain increasing interest making powerful model understandable interpretable narrowing cause good bad prediction focus recurrent neural network RNNs state art mo...
[-0.003315289504826069, 0.03745390102267265, -0.001666680327616632, 0.03461119160056114, -0.0002736737369559705, -0.008985842578113079, 0.013043899089097977, 0.02389092929661274, -0.05586247146129608, -0.030181625857949257, -0.015812372788786888, 0.001096821273677051, 0.0483892448246479, 0.03914790228009224, -0.0120239...
1,779
1,779
['Vikrant Singh Tomar', 'Richard C. Rose']
1606.05925v1
Deep neural networks (DNNs) have been successfully applied to a wide variety of acoustic modeling tasks in recent years. These include the applications of DNNs either in a discriminative feature extraction or in a hybrid acoustic modeling scenario. Despite the rapid progress in this area, a number of challenges remain ...
Graph based manifold regularized deep neural networks for automatic speech recognition
2,016
http://arxiv.org/pdf/1606.05925v1
Title Graph based manifold regularized deep neural network automatic speech recognition Summary Deep neural network DNNs successfully applied wide variety acoustic modeling task recent year include application DNNs either discriminative feature extraction hybrid acoustic modeling scenario Despite rapid progress area nu...
[-0.024200448766350746, -0.007004876621067524, 0.010889663361012936, 0.059166375547647476, -0.012513612397015095, -0.02935037761926651, 0.037132080644369125, -0.0114065483212471, 0.05420541390776634, 0.014456327073276043, -0.03439721465110779, 0.03579207509756088, 0.012090773321688175, 0.02287927083671093, 0.0656088814...
1,780
1,780
['Tolga Bolukbasi', 'Kai-Wei Chang', 'James Zou', 'Venkatesh Saligrama', 'Adam Kalai']
1606.06121v1
Machine learning algorithms are optimized to model statistical properties of the training data. If the input data reflects stereotypes and biases of the broader society, then the output of the learning algorithm also captures these stereotypes. In this paper, we initiate the study of gender stereotypes in {\em word emb...
Quantifying and Reducing Stereotypes in Word Embeddings
2,016
http://arxiv.org/pdf/1606.06121v1
Title Quantifying Reducing Stereotypes Word Embeddings Summary Machine learning algorithm optimized model statistical property training data input data reflects stereotype bias broader society output learning algorithm also capture stereotype paper initiate study gender stereotype em word embedding popular framework re...
[0.03147835657000542, 0.08870688080787659, -0.025211628526449203, 0.009187771938741207, 0.025265010073781013, 0.03471821919083595, 0.021053830161690712, -0.02742585353553295, 0.00656746793538332, -0.05600842833518982, 0.023303212597966194, -0.01625075191259384, 0.06256075203418732, 0.021036772057414055, 0.0240676756948...
1,781
1,781
['Abram Handler', 'Su Lin Blodgett', "Brendan O'Connor"]
1606.06352v1
We explore two techniques which use color to make sense of statistical text models. One method uses in-text annotations to illustrate a model's view of particular tokens in particular documents. Another uses a high-level, "words-as-pixels" graphic to display an entire corpus. Together, these methods offer both zoomed-i...
Visualizing textual models with in-text and word-as-pixel highlighting
2,016
http://arxiv.org/pdf/1606.06352v1
Title Visualizing textual model intext wordaspixel highlighting Summary explore two technique use color make sense statistical text model One method us intext annotation illustrate model view particular token particular document Another us highlevel wordsaspixels graphic display entire corpus Together method offer zoom...
[0.023736784234642982, 0.021046586334705353, -0.005009775515645742, 0.03973695635795593, -0.04097811132669449, 0.01023758202791214, 0.02307870239019394, 0.052063554525375366, -0.01909453235566616, -0.0699344277381897, 0.03782320022583008, 0.025394702330231667, 0.024955566972494125, 0.0655747503042221, -0.02318382449448...
1,782
1,782
['Abulhair Saparov', 'Tom M. Mitchell']
1606.06361v1
We present a framework that couples the syntax and semantics of natural language sentences in a generative model, in order to develop a semantic parser that jointly infers the syntactic, morphological, and semantic representations of a given sentence under the guidance of background knowledge. To generate a sentence in...
A Probabilistic Generative Grammar for Semantic Parsing
2,016
http://arxiv.org/pdf/1606.06361v1
Title Probabilistic Generative Grammar Semantic Parsing Summary present framework couple syntax semantics natural language sentence generative model order develop semantic parser jointly infers syntactic morphological semantic representation given sentence guidance background knowledge generate sentence framework seman...
[0.03659990057349205, 0.03879448398947716, 0.008073045872151852, 0.012865204364061356, -0.03769800812005997, 0.02284981869161129, -0.005859780590981245, -0.02666627988219261, -0.02617436833679676, -0.08894909918308258, 0.04152030870318413, 0.03657611086964607, -0.019227463752031326, 0.04805707186460495, 0.0089739654213...
1,783
1,783
['Kyle Reing', 'David C. Kale', 'Greg Ver Steeg', 'Aram Galstyan']
1606.07043v1
Many predictive tasks, such as diagnosing a patient based on their medical chart, are ultimately defined by the decisions of human experts. Unfortunately, encoding experts' knowledge is often time consuming and expensive. We propose a simple way to use fuzzy and informal knowledge from experts to guide discovery of int...
Toward Interpretable Topic Discovery via Anchored Correlation Explanation
2,016
http://arxiv.org/pdf/1606.07043v1
Title Toward Interpretable Topic Discovery via Anchored Correlation Explanation Summary Many predictive task diagnosing patient based medical chart ultimately defined decision human expert Unfortunately encoding expert knowledge often time consuming expensive propose simple way use fuzzy informal knowledge expert guide...
[0.0270132627338171, 0.030574867501854897, -0.03197349235415459, -0.027627047151327133, 0.0009772945195436478, 0.0024050287902355194, -0.016540024429559708, 0.04842615872621536, -0.033170491456985474, 0.0443897508084774, 0.018261820077896118, 0.04414881020784378, 0.03603944182395935, 0.04836519435048103, -0.04039026051...
1,784
1,784
['Tsendsuren Munkhdalai', 'Hong Yu']
1607.04315v3
We present a memory augmented neural network for natural language understanding: Neural Semantic Encoders. NSE is equipped with a novel memory update rule and has a variable sized encoding memory that evolves over time and maintains the understanding of input sequences through read}, compose and write operations. NSE c...
Neural Semantic Encoders
2,016
http://arxiv.org/pdf/1607.04315v3
Title Neural Semantic Encoders Summary present memory augmented neural network natural language understanding Neural Semantic Encoders NSE equipped novel memory update rule variable sized encoding memory evolves time maintains understanding input sequence read compose write operation NSE also access multiple shared mem...
[0.04707041382789612, 0.00234998669475317, 0.0035761974286288023, 0.0696980357170105, 3.4723325370578095e-06, 0.01351860910654068, 0.00972059927880764, -0.028394779190421104, -0.025753699243068695, -0.0663306787610054, -0.009739781729876995, 0.00723453052341938, 0.008532784879207611, 0.0580628402531147, 0.0186619609594...
1,785
1,785
['Tsendsuren Munkhdalai', 'Hong Yu']
1607.04492v2
Recurrent neural networks (RNNs) process input text sequentially and model the conditional transition between word tokens. In contrast, the advantages of recursive networks include that they explicitly model the compositionality and the recursive structure of natural language. However, the current recursive architectur...
Neural Tree Indexers for Text Understanding
2,016
http://arxiv.org/pdf/1607.04492v2
Title Neural Tree Indexers Text Understanding Summary Recurrent neural network RNNs process input text sequentially model conditional transition word token contrast advantage recursive network include explicitly model compositionality recursive structure natural language However current recursive architecture limited d...
[0.056137993931770325, 0.0014515712391585112, -0.008718484081327915, 0.061887480318546295, -0.06320414692163467, 0.01116278674453497, -0.028818244114518166, -0.009448040276765823, -0.004465483129024506, -0.05766511335968971, 0.01874983124434948, -0.0019493388244882226, 0.0054349289275705814, 0.0504547581076622, -0.0105...
1,786
1,786
['Khanh Nguyen']
1607.05241v1
We present a novel view that unifies two frameworks that aim to solve sequential prediction problems: learning to search (L2S) and recurrent neural networks (RNN). We point out equivalences between elements of the two frameworks. By complementing what is missing from one framework comparing to the other, we introduce a...
Imitation Learning with Recurrent Neural Networks
2,016
http://arxiv.org/pdf/1607.05241v1
Title Imitation Learning Recurrent Neural Networks Summary present novel view unifies two framework aim solve sequential prediction problem learning search L2S recurrent neural network RNN point equivalence element two framework complementing missing one framework comparing introduce advanced imitation learning framewo...
[0.020444855093955994, -0.008145631290972233, 0.015189497731626034, 0.012216116301715374, -0.0036221712362021208, -0.011081886477768421, -0.019351588562130928, -0.00746134715154767, 0.00034302985295653343, -0.0022916009183973074, -0.02786164917051792, -0.029778001829981804, 0.017808590084314346, -0.010111356154084206, ...
1,787
1,787
['Jiazhen He', 'Benjamin I. P. Rubinstein', 'James Bailey', 'Rui Zhang', 'Sandra Milligan']
1607.08720v2
This paper explores the suitability of using automatically discovered topics from MOOC discussion forums for modelling students' academic abilities. The Rasch model from psychometrics is a popular generative probabilistic model that relates latent student skill, latent item difficulty, and observed student-item respons...
TopicResponse: A Marriage of Topic Modelling and Rasch Modelling for Automatic Measurement in MOOCs
2,016
http://arxiv.org/pdf/1607.08720v2
Title TopicResponse Marriage Topic Modelling Rasch Modelling Automatic Measurement MOOCs Summary paper explores suitability using automatically discovered topic MOOC discussion forum modelling student academic ability Rasch model psychometrics popular generative probabilistic model relates latent student skill latent i...
[0.04231343790888786, 0.0014509187312796712, -0.03321237117052078, 0.008962738327682018, -0.004874985199421644, 0.01998598873615265, 0.02259245328605175, -0.044399525970220566, -0.021258708089590073, -0.041155241429805756, 0.013322918675839901, 0.03498445823788643, -0.025366326794028282, 0.047182586044073105, -0.032685...
1,788
1,788
['Rie Johnson', 'Tong Zhang']
1609.00718v1
This paper reports the performances of shallow word-level convolutional neural networks (CNN), our earlier work (2015), on the eight datasets with relatively large training data that were used for testing the very deep character-level CNN in Conneau et al. (2016). Our findings are as follows. The shallow word-level CNN...
Convolutional Neural Networks for Text Categorization: Shallow Word-level vs. Deep Character-level
2,016
http://arxiv.org/pdf/1609.00718v1
Title Convolutional Neural Networks Text Categorization Shallow Wordlevel v Deep Characterlevel Summary paper report performance shallow wordlevel convolutional neural network CNN earlier work 2015 eight datasets relatively large training data used testing deep characterlevel CNN Conneau et al 2016 finding follows shal...
[0.056918513029813766, 0.058057140558958054, 0.017407378181815147, 0.026695996522903442, -0.03327189013361931, 0.03285828232765198, 0.07661614567041397, 0.03129253163933754, -0.014548549428582191, -0.08121126145124435, -0.019220495596528053, -0.031023792922496796, -0.001981391804292798, 0.022541742771863937, -0.0073837...
1,789
1,789
['Trapit Bansal', 'David Belanger', 'Andrew McCallum']
1609.02116v2
In a variety of application domains the content to be recommended to users is associated with text. This includes research papers, movies with associated plot summaries, news articles, blog posts, etc. Recommendation approaches based on latent factor models can be extended naturally to leverage text by employing an exp...
Ask the GRU: Multi-Task Learning for Deep Text Recommendations
2,016
http://arxiv.org/pdf/1609.02116v2
Title Ask GRU MultiTask Learning Deep Text Recommendations Summary variety application domain content recommended user associated text includes research paper movie associated plot summary news article blog post etc Recommendation approach based latent factor model extended naturally leverage text employing explicit ma...
[0.05633292347192764, 0.013745589181780815, -0.0005606191698461771, 0.025181276723742485, 0.011170759797096252, -0.008084817789494991, 0.028113894164562225, 0.0029687387868762016, -0.01215443667024374, -0.035419829189777374, -0.06382983177900314, -0.008363747037947178, 0.009878422133624554, 0.1118076965212822, 0.001676...
1,790
1,790
['Kar Wai Lim', 'Wray Buntine', 'Changyou Chen', 'Lan Du']
1609.06783v1
The Dirichlet process and its extension, the Pitman-Yor process, are stochastic processes that take probability distributions as a parameter. These processes can be stacked up to form a hierarchical nonparametric Bayesian model. In this article, we present efficient methods for the use of these processes in this hierar...
Nonparametric Bayesian Topic Modelling with the Hierarchical Pitman-Yor Processes
2,016
http://arxiv.org/pdf/1609.06783v1
Title Nonparametric Bayesian Topic Modelling Hierarchical PitmanYor Processes Summary Dirichlet process extension PitmanYor process stochastic process take probability distribution parameter process stacked form hierarchical nonparametric Bayesian model article present efficient method use process hierarchical context ...
[0.017766788601875305, 0.013873778283596039, -0.03145155310630798, -0.03713389113545418, -0.03500792756676674, -8.573859668103978e-05, 0.00411535007879138, -0.006711432244628668, -0.06483971327543259, -0.03998079150915146, 0.0032539847306907177, 0.02304503135383129, 0.020068254321813583, 0.035275667905807495, -0.004127...
1,791
1,791
['Akash Kumar Dhaka', 'Giampiero Salvi']
1610.00520v1
We propose the application of a semi-supervised learning method to improve the performance of acoustic modelling for automatic speech recognition based on deep neural net- works. As opposed to unsupervised initialisation followed by supervised fine tuning, our method takes advantage of both unlabelled and labelled data...
Semi-supervised Learning with Sparse Autoencoders in Phone Classification
2,016
http://arxiv.org/pdf/1610.00520v1
Title Semisupervised Learning Sparse Autoencoders Phone Classification Summary propose application semisupervised learning method improve performance acoustic modelling automatic speech recognition based deep neural net work opposed unsupervised initialisation followed supervised fine tuning method take advantage unlab...
[-0.00911770574748516, 0.039553165435791016, -0.0013491939753293991, 0.035480331629514694, 0.040480051189661026, -0.007923977449536324, 0.08867798745632172, -0.0337616503238678, 0.0037770934868603945, -0.020854493603110313, -0.0512445792555809, -0.022023100405931473, 0.015872718766331673, 0.06759782880544662, 0.0226307...
1,792
1,792
['William Chan', 'Yu Zhang', 'Quoc Le', 'Navdeep Jaitly']
1610.03035v6
We present the Latent Sequence Decompositions (LSD) framework. LSD decomposes sequences with variable lengthed output units as a function of both the input sequence and the output sequence. We present a training algorithm which samples valid extensions and an approximate decoding algorithm. We experiment with the Wall ...
Latent Sequence Decompositions
2,016
http://arxiv.org/pdf/1610.03035v6
Title Latent Sequence Decompositions Summary present Latent Sequence Decompositions LSD framework LSD decomposes sequence variable lengthed output unit function input sequence output sequence present training algorithm sample valid extension approximate decoding algorithm experiment Wall Street Journal speech recogniti...
[-0.010576996020972729, 0.08419755846261978, 0.007259641773998737, 0.02302272990345955, 0.007892245426774025, 0.00026789383264258504, 0.01651039905846119, 0.005796636920422316, -0.054970040917396545, -0.018974576145410538, 0.013920764438807964, -0.013442758470773697, 0.06122255697846413, 0.051456063985824585, -0.002923...
1,793
1,793
['Yacine Jernite', 'Anna Choromanska', 'David Sontag']
1610.04658v2
We consider multi-class classification where the predictor has a hierarchical structure that allows for a very large number of labels both at train and test time. The predictive power of such models can heavily depend on the structure of the tree, and although past work showed how to learn the tree structure, it expect...
Simultaneous Learning of Trees and Representations for Extreme Classification and Density Estimation
2,016
http://arxiv.org/pdf/1610.04658v2
Title Simultaneous Learning Trees Representations Extreme Classification Density Estimation Summary consider multiclass classification predictor hierarchical structure allows large number label train test time predictive power model heavily depend structure tree although past work showed learn tree structure expected f...
[0.014153022319078445, 0.0050354390405118465, -0.019722996279597282, 0.024005791172385216, -0.0029259573202580214, -0.011855348944664001, -0.01858096942305565, 0.019322017207741737, -0.025439253076910973, -0.06801143288612366, -0.02276722528040409, -0.006928304210305214, -0.014313033781945705, 0.05205293372273445, -0.0...
1,794
1,794
['Hao Tang', 'Weiran Wang', 'Kevin Gimpel', 'Karen Livescu']
1610.06700v1
Recent work on discriminative segmental models has shown that they can achieve competitive speech recognition performance, using features based on deep neural frame classifiers. However, segmental models can be more challenging to train than standard frame-based approaches. While some segmental models have been success...
End-to-End Training Approaches for Discriminative Segmental Models
2,016
http://arxiv.org/pdf/1610.06700v1
Title EndtoEnd Training Approaches Discriminative Segmental Models Summary Recent work discriminative segmental model shown achieve competitive speech recognition performance using feature based deep neural frame classifier However segmental model challenging train standard framebased approach segmental model successfu...
[-0.028478937223553658, 0.016540316864848137, 0.021432047709822655, 0.06432463228702545, 0.008129404857754707, 0.002682165242731571, 0.0377582348883152, -0.011096011847257614, -0.07042338699102402, -0.015276467427611351, -0.04922160878777504, -0.04419124871492386, 0.008410793729126453, 0.04700691998004913, -0.009704271...
1,795
1,795
['Jiaqi Mu', 'Suma Bhat', 'Pramod Viswanath']
1610.07569v1
Vector representations of words have heralded a transformational approach to classical problems in NLP; the most popular example is word2vec. However, a single vector does not suffice to model the polysemous nature of many (frequent) words, i.e., words with multiple meanings. In this paper, we propose a three-fold appr...
Geometry of Polysemy
2,016
http://arxiv.org/pdf/1610.07569v1
Title Geometry Polysemy Summary Vector representation word heralded transformational approach classical problem NLP popular example word2vec However single vector suffice model polysemous nature many frequent word ie word multiple meaning paper propose threefold approach unsupervised polysemy modeling context represent...
[0.04916173592209816, -0.0010517311748117208, -0.03140999376773834, 0.09193380922079086, -0.04143262654542923, -0.007557207718491554, -0.009549979120492935, 0.03270125389099121, -0.013072668574750423, -0.08348380029201508, 0.036605916917324066, 0.02308306097984314, 0.05181964859366417, 0.002037076512351632, 0.015122766...
1,796
1,796
['Hakan Inan', 'Khashayar Khosravi', 'Richard Socher']
1611.01462v3
Recurrent neural networks have been very successful at predicting sequences of words in tasks such as language modeling. However, all such models are based on the conventional classification framework, where the model is trained against one-hot targets, and each word is represented both as an input and as an output in ...
Tying Word Vectors and Word Classifiers: A Loss Framework for Language Modeling
2,016
http://arxiv.org/pdf/1611.01462v3
Title Tying Word Vectors Word Classifiers Loss Framework Language Modeling Summary Recurrent neural network successful predicting sequence word task language modeling However model based conventional classification framework model trained onehot target word represented input output isolation cause inefficiency learning...
[0.03282519429922104, 0.03031608834862709, -0.003166421316564083, 0.02939845621585846, -0.01246609166264534, -0.0067560700699687, 0.009748129174113274, 0.007715396583080292, -0.05070029944181442, -0.06225309148430824, 0.019687151536345482, -0.036549899727106094, 0.0690421313047409, 0.05903236195445061, 0.04410680755972...
1,797
1,797
['Tsubasa Ochiai', 'Shigeki Matsuda', 'Hideyuki Watanabe', 'Shigeru Katagiri']
1611.05527v1
We examine the effect of the Group Lasso (gLasso) regularizer in selecting the salient nodes of Deep Neural Network (DNN) hidden layers by applying a DNN-HMM hybrid speech recognizer to TED Talks speech data. We test two types of gLasso regularization, one for outgoing weight vectors and another for incoming weight vec...
Automatic Node Selection for Deep Neural Networks using Group Lasso Regularization
2,016
http://arxiv.org/pdf/1611.05527v1
Title Automatic Node Selection Deep Neural Networks using Group Lasso Regularization Summary examine effect Group Lasso gLasso regularizer selecting salient node Deep Neural Network DNN hidden layer applying DNNHMM hybrid speech recognizer TED Talks speech data test two type gLasso regularization one outgoing weight ve...
[0.012201865203678608, 0.03162243217229843, -0.010533126071095467, 0.025301536545157433, 0.01792096719145775, -0.01995651051402092, 0.08143311738967896, 0.0008843312971293926, 0.016202442348003387, 0.012774349190294743, -0.030036579817533493, 0.0749451220035553, 0.041618797928094864, -0.033619608730077744, 0.0363608673...
1,798
1,798
['Jacob Eisenstein']
1611.06933v1
In lexicon-based classification, documents are assigned labels by comparing the number of words that appear from two opposed lexicons, such as positive and negative sentiment. Creating such words lists is often easier than labeling instances, and they can be debugged by non-experts if classification performance is unsa...
Unsupervised Learning for Lexicon-Based Classification
2,016
http://arxiv.org/pdf/1611.06933v1
Title Unsupervised Learning LexiconBased Classification Summary lexiconbased classification document assigned label comparing number word appear two opposed lexicon positive negative sentiment Creating word list often easier labeling instance debugged nonexperts classification performance unsatisfactory However little ...
[0.050820544362068176, 0.04211849346756935, 0.0029872660525143147, 0.05428757146000862, -0.05543491989374161, 0.006732850801199675, 0.031127236783504486, 0.007037898059934378, 0.013284382410347462, -0.06434458494186401, -0.0013391758548095822, 0.021669834852218628, 0.013461035676300526, 0.0344788059592247, -0.029217256...
1,799
1,799
['Raghavendra Chalapathy', 'Ehsan Zare Borzeshi', 'Massimo Piccardi']
1611.08373v1
Automated extraction of concepts from patient clinical records is an essential facilitator of clinical research. For this reason, the 2010 i2b2/VA Natural Language Processing Challenges for Clinical Records introduced a concept extraction task aimed at identifying and classifying concepts into predefined categories (i....
Bidirectional LSTM-CRF for Clinical Concept Extraction
2,016
http://arxiv.org/pdf/1611.08373v1
Title Bidirectional LSTMCRF Clinical Concept Extraction Summary Automated extraction concept patient clinical record essential facilitator clinical research reason 2010 i2b2VA Natural Language Processing Challenges Clinical Records introduced concept extraction task aimed identifying classifying concept predefined cate...
[0.06030861288309097, 0.013960923999547958, 0.02593321166932583, 0.040036290884017944, -0.01722150295972824, 0.021450858563184738, -0.031128741800785065, 0.03607649356126785, -0.042241502553224564, -0.031446680426597595, 0.004907954018563032, -0.07109315693378448, 0.016512200236320496, 0.0751417875289917, 0.00206499174...