01524nas a2200193 4500008004100000245005600041210005500097300001200152520098200164653002701146653002301173653001701196100001601213700001301229700001401242700001801256700001601274856004001290 2012 eng d00aDiscovering Fine-grained Sentiment in Suicide Notes0 aDiscovering Finegrained Sentiment in Suicide Notes a137-1453 aThis paper presents our solution for the i2b2 sentiment classification challenge. Our hybrid system consists of machine learning and rule-based classifiers. For the machine learning classifier, we investigate a variety of lexical, syntactic and knowledge-based features, and show how much these features contribute to the performance of the classifier through experiments. For the rule-based classifier, we propose an algorithm to automatically extract effective syntactic and lexical patterns from training examples. The experimental results show that the rule-based classifier outperforms the baseline machine learning classifier using unigram features. By combining the machine learning classifier and the rule-based classifier, the hybrid system gains a better trade-off between precision and recall, and yields the highest micro-averaged F-measure (0.5038), which is better than the mean (0.4875) and median (0.5027) micro-average F-measures among all participating teams.10aEmotion Identification10aSentiment Analysis10aSuicide Note1 aWang, Wenbo1 aChen, Lu1 aTan, Ming1 aWang, Shaojun1 aSheth, Amit uhttp://knoesis.wright.edu/node/159102230nas a2200241 4500008004100000245008900041210006900130260011300199520140800312653001401720653001901734653001701753653002301770653002501793653002501818653001701843100001301860700001601873700002501889700001801914700001601932856004001948 2012 eng d00aExtracting Diverse Sentiment Expressions with Target-dependent Polarity from Twitter0 aExtracting Diverse Sentiment Expressions with Targetdependent Po aDublin, IrelandbIn Proceedings of the 6th International AAAI Conference on Weblogs and Social Media (ICWSM)3 aThe problem of automatic extraction of sentiment expressions from informal text, as in microblogs such as tweets is a recent area of investigation. Compared to formal text, such as in product reviews or news articles, one of the key challenges lies in the wide diversity and informal nature of sentiment expressions that cannot be trivially enumerated or captured using predefined lexical patterns. In this work, we present an optimization-based approach to automatically extract sentiment expressions for a given target (e.g., movie, or person) from a corpus of unlabeled tweets. Specifically, we make three contributions: (i) we recognize a diverse and richer set of sentiment-bearing expressions in tweets, including formal and slang words/phrases, not limited to pre-specified syntactic patterns; (ii) instead of associating sentiment with an entire tweet, we assess the target-dependent polarity of each sentiment expression. The polarity of sentiment expression is determined by the nature of its target; (iii) we provide a novel formulation of assigning polarity to a sentiment expression as a constrained optimization problem over the tweet corpus. Experiments conducted on two domains, tweets mentioning movie and person entities, show that our approach improves accuracy in comparison with several baseline methods, and that the improvement becomes more prominent with increasing corpus sizes.10aMicroblog10aOpinion Mining10aOptimization10aSentiment Analysis10aSentiment Expression10aSentiment Extraction10aSocial Media1 aChen, Lu1 aWang, Wenbo1 aNagarajan, Meenakshi1 aWang, Shaojun1 aSheth, Amit uhttp://knoesis.wright.edu/node/120702655nas a2200157 4500008004100000245010100041210006900142520207400211653008402285100001302369700001602382700002502398700001802423700001602441856004002457 2011 eng d00aBeyond Positive/Negative Classification: Automatic Extraction of Sentiment Clues from Microblogs0 aBeyond PositiveNegative Classification Automatic Extraction of S3 aMicroblogging provides a large volume of text for learning and understanding people's sentiments on a variety of topics. Much of the current work on sentiment analysis of microblogs (e.g., tweets) focuses on document level polarity. However, identifying sentiment clues with respect to specific targets (e.g., named entities) can be more useful than pure document polarity results. For example, sentiment clues such as 'must see', 'awesome', 'rate 5 stars' (in the movie domain) are much more meaningful than the polarities of tweets only. Previous attempts at single-word sentiment clue extraction from formal text will not suffice for extracting multi-word sentiment phrases. Single words 'must' and 'see' do not separately convey polarity, but their combination 'must see' expresses strong positive sentiment towards a movie target. Another issue with identifying sentiment clues is identifying informal sentiment expressions, such as misspellings ('kool'), abbreviations ('wtf') and slangs ('da bomb'). In this paper, we propose an approach for automatically extracting both single-word and multi-word sentiment clues. Such clues can include both traditional and slang expressions. We also present a mechanism for assessing their target-specific polarities from an unlabeled microblog corpus. Our approach first leverages traditional and slang subjective lexicons to generate candidate sentiment clues given some specific target. It then incorporates inter-clue relations from corpora into an optimization model to estimate the probability of a clue denoting positive/negative sentiment. Experiments using microblog data sets on two different domains -- movie and person -- show that the proposed approach can effectively 1) extract single-word as well as phrase sentiment clues, 2) identify both traditional and slang sentiment clues, and 3) determine their target-specific polarities. We also demonstrate how the proposed approach is superior in comparison with several baseline methods.10aOptimization and Opinion Mining and Sentiment Analysis and Sentiment Extraction1 aChen, Lu1 aWang, Wenbo1 aNagarajan, Meenakshi1 aWang, Shaojun1 aSheth, Amit uhttp://knoesis.wright.edu/node/191001450nas a2200157 4500008004100000245007700041210006900118260005400187520092200241653002901163100001601192700001301208700001801221700001301239856004001252 2010 eng d00aA Rate Distortion Approach for Semi-Supervised Conditional Random Fields0 aRate Distortion Approach for SemiSupervised Conditional Random F bAdvances in Neural Information Processing Systems3 aWe propose a novel information theoretic approach for semi-supervised learning of conditional random fields that defines a training objective to combine the conditional likelihood on labeled data and the mutual information on unlabeled data. In contrast to previous minimum conditional entropy semi-supervised discriminative learning methods, our approach is grounded on a more solid foundation, the rate distortion theory in information theory. We analyze the tractability of the framework for structured prediction and present a convergent variational training algorithm to defy the combinatorial explosion of terms in the sum over label configurations. Our experimental results show the rate distortion approach outperforms standard l2 regularization, minimum conditional entropy regularization as well as maximum conditional entropy regularization on both multi-class classifcation and sequence labeling problems.10asemi-supervised learning1 aHaffari, G.1 aWang, Y.1 aWang, Shaojun1 aMori, G. uhttp://knoesis.wright.edu/node/105801220nas a2200133 4500008004100000245007000041210006900111260001800180520080200198100001501000700001301015700001801028856004001046 2009 eng d00aInformation Theoretic Regularization for Semi-Supervised Boosting0 aInformation Theoretic Regularization for SemiSupervised Boosting aParis, France3 aWe present novel semi-supervised boosting algorithms that incrementally build linear combinations of weak classifiers through generic functional gradient descent using both labeled and unlabeled training data. Our approach is based on extending information regularization framework to boosting,bearing loss functions that combine log loss on labeled data with the information-theoretic measures to encode unlabeled data. Even though the information-theoretic regularization terms make the optimization non-convex, we propose simple sequential gradient descent optimization algorithms, and obtain impressively improved results on synthetic, benchmark and real world tasks over supervised boosting algorithms which use the labeled data alone and a state-of-the-art semi-supervised boosting algorithm1 aZheng, Lei1 aLiu, Yan1 aWang, Shaojun uhttp://knoesis.wright.edu/node/105700424nas a2200121 4500008004100000245007700041210006900118100001800187700002500205700001600230700001600246856004000262 2009 eng d00aMonetizing User Activity on Social Networks - Challenges and Experiences0 aMonetizing User Activity on Social Networks Challenges and Exper1 aWang, Shaojun1 aNagarajan, Meenakshi1 aBaid, Kamal1 aSheth, Amit uhttp://knoesis.wright.edu/node/112300511nas a2200133 4500008004100000245007700041210006900118260007500187100002500262700001600287700001600303700001800319856004000337 2009 eng d00aMonetizing User Activity on Social Networks - Challenges and Experiences0 aMonetizing User Activity on Social Networks Challenges and Exper bBeyond Search: Semantic Computing and Internet Economics 2009 Workshop1 aNagarajan, Meenakshi1 aBaid, Kamal1 aSheth, Amit1 aWang, Shaojun uhttp://knoesis.wright.edu/node/181800370nas a2200133 4500008004100000245004100041210004100082100001800123700001600141700001300157700001300170700001300183856004000196 2008 eng d00aBoosting with Incomplete Information0 aBoosting with Incomplete Information1 aWang, Shaojun1 aHaffari, G.1 aJiao, F.1 aWang, Y.1 aMori, G. uhttp://knoesis.wright.edu/node/112500390nas a2200133 4500008004100000245005000041210005000091100001200141700001800153700001400171700001500185700001600200856004000216 2008 eng d00aConstrained Classification on Structured Data0 aConstrained Classification on Structured Data1 aLee, C.1 aWang, Shaojun1 aBrown, M.1 aMurtha, A.1 aGreiner, R. uhttp://knoesis.wright.edu/node/111001222nas a2200145 4500008004100000245004800041210004800089520070700137653011700844100001600961700002500977700001601002700001801018856004001036 2008 eng d00aMonetizing User Activity on Social Networks0 aMonetizing User Activity on Social Networks3 aIn this work, we investigate techniques to monitize user activity on public forums, marketplaces and groups on social network sites. Our approach involves (a) identifying the monetization potential of user posts and (b) eliminating o_- topic content in monetizable posts to use the most relevant keywords for advertising. Our _rst user study involving 30 users and data from MySpace and Facebook, shows that 52% of ad impressions shown after using our system were more targeted compared to the 30% relevant impressions generated without using our system. A second smaller study suggests that pro_le ads that are based on user activity generate more interest than ads solely based on pro_le information.10aSocial Networks and Monetization and User activity and Computational Advertising and O-topic content and Intents1 aSheth, Amit1 aNagarajan, Meenakshi1 aBaid, Kamal1 aWang, Shaojun uhttp://knoesis.wright.edu/node/193700423nas a2200133 4500008004100000245006700041210006600108100001400174700001500188700001200203700001800215700001600233856004000249 2008 eng d00aSegmenting Brain Tumors Using Pseudo-Conditional Random Fields0 aSegmenting Brain Tumors Using PseudoConditional Random Fields1 aBrown, M.1 aMurtha, A.1 aLee, C.1 aWang, Shaojun1 aGreiner, R. uhttp://knoesis.wright.edu/node/110901051nas a2200145 4500008004100000245005500041210005500096520053500151653010400686100001600790700002500806700001600831700001800847856004000865 2008 eng d00aTargeted Content Delivery for Social Media Content0 aTargeted Content Delivery for Social Media Content3 aSpotting contextually relevant keywords is fundamental to effective content suggestions on the Web. In this regard, misspellings, entity variations and off-topic discussions in content from Social Media pose unique challenges. Here, we present an algorithm that assists content delivery systems by identifying contextually relevant keywords and eliminating off-topic keywords. A preliminary user study over data from MySpace and Facebook clearly suggests the usefulness of our work in delivering more targeted content suggestions.10aMutual Information and Contextual keywords and Contextual Content Delivery and Social Media Content1 aSheth, Amit1 aNagarajan, Meenakshi1 aBaid, Kamal1 aWang, Shaojun uhttp://knoesis.wright.edu/node/196101307nas a2200133 4500008004100000245007600041210006900117520086600186100001801052700002501070700002201095700001601117856004001133 2008 eng d00aUnsupervised Discovery of Compound Entities for Relationship Extraction0 aUnsupervised Discovery of Compound Entities for Relationship Ext3 aIn this paper we investigate unsupervised population of a biomedical ontology via information extraction from biomedical literature. Relationships in text seldom connect simple entities. We therefore focus on identifying compound entities rather than mentions of simple entities. We present a method based on rules over grammatical dependency structures for unsupervised segmentation of sentences into compound entities and relationships. We complement the rule-based approach with a statistical component that prunes structures with low information content, thereby reducing false positives in the prediction of compound entities, their constituents and relationships. The extraction is manually evaluated with respect to the UMLS Semantic Network by analyzing the conformance of the extracted triples with the corresponding UMLS relationship type definitions.1 aWang, Shaojun1 aRamakrishnan, Cartic1 aMendes, Pablo, N.1 aSheth, Amit uhttp://knoesis.wright.edu/node/103901037nas a2200145 4500008004100000245004200041210004200083520063600125100001400761700002100775700001900796700001800815700001800833856004000851 2007 eng d00aImplicit Online Learning with Kernels0 aImplicit Online Learning with Kernels3 aWe present two new algorithms for online learning in reproducing kernel Hilbert spaces. Our first algorithm, ILK (implicit online learning with kernels), employs a new, implicit update technique that can be applied to a wide variety of convex loss functions. We then introduce a bounded memory version, SILK (sparse ILK), that maintains a compact representation of the predictor without compromising solution quality, even in non-stationary environments. We prove loss bounds and analyze the convergence rate of both. Experimental evidence shows that our proposed algorithms outperform current methods on synthetic and real data.
1 aCheng, L.1 aVishwanathan, S.1 aSchuurmans, D.1 aWang, Shaojun1 aCaelli, Terry uhttp://knoesis.wright.edu/node/145000449nas a2200133 4500008004100000245008700041210006900128100001300197700001900210700001600229700001200245700001800257856004000275 2007 eng d00aLearning to Model Spatial Dependency: Semi-Supervised Discriminative Random Fields0 aLearning to Model Spatial Dependency SemiSupervised Discriminati1 aJiao, F.1 aSchuurmans, D.1 aGreiner, R.1 aLee, C.1 aWang, Shaojun uhttp://knoesis.wright.edu/node/147200904nas a2200121 4500008004100000245009600041210006900137300001400206520049100220100001300711700001800724856004000742 2006 eng d00aAlmost Sure Convergence of Titterington's Recursive Estimator for Finite Mixture Models0 aAlmost Sure Convergence of Titterington39s Recursive Estimator f a2001-10063 aTitterington proposed a recursive parameter estimation algorithm for finite mixture models. However, due to the well known problem of singularities and multiple maximum, minimum and saddle points that are possible on the likelihood surfaces, convergence analysis has seldom been made in the past years. In this paper, under mild conditions, we show the global convergence of Titterington's recursive estimator and its MAP variant for mixture models of full regular exponential family.
1 aZhao, Y.1 aWang, Shaojun uhttp://knoesis.wright.edu/node/145500898nas a2200121 4500008004100000245009200041210006900133300001400202520048900216100001800705700001300723856004000736 2006 eng d00aAlmost Sure Convergence of Titterington's Recursive Estimator for Finite Mixture Models0 aAlmost Sure Convergence of Titteringtons Recursive Estimator for a2001-20063 aTitterington proposed a recursive parameter estimation algorithm for finite mixture models. However, due to the well known problem of singularities and multiple maximum, minimum and saddle points that are possible on the likelihood surfaces, convergence analysis has seldom been made in the past years. In this paper, under mild conditions, we show the global convergence of Titterington's recursive estimator and its MAP variant for mixture models of full regular exponential family.1 aWang, Shaojun1 aZhao, Y. uhttp://knoesis.wright.edu/node/240800427nas a2200133 4500008004100000245006400041210006100105100002100166700001500187700001400202700001900216700001800235856004000253 2006 eng d00aAn Online Discriminative Approach to Background Subtraction0 aOnline Discriminative Approach to Background Subtraction1 aVishwanathan, S.1 aCaelli, T.1 aCheng, L.1 aSchuurmans, D.1 aWang, Shaojun uhttp://knoesis.wright.edu/node/114400456nas a2200133 4500008004100000245009400041210006900135100001800204700001600222700001300238700001900251700001200270856004000282 2006 eng d00aSemi-Supervised Conditional Random Fields for Improved Sequence Segmentation and Labeling0 aSemiSupervised Conditional Random Fields for Improved Sequence S1 aWang, Shaojun1 aGreiner, R.1 aJiao, F.1 aSchuurmans, D.1 aLee, C. uhttp://knoesis.wright.edu/node/106600488nas a2200145 4500008004100000245008300041210006900124260001700193100001800210700001800228700001400246700002100260700002100281856004000302 2006 eng d00aStochastic Analysis of Lexical and Semantic Enhanced Structural Language Model0 aStochastic Analysis of Lexical and Semantic Enhanced Structural aTokyo, Japan1 aWang, Shaojun1 aWang, Shaomin1 aCheng, Li1 aGreiner, Russell1 aSchuurmans, Dale uhttp://knoesis.wright.edu/node/107200380nas a2200109 4500008004100000245007400041210006900115100001600184700001800200700001200218856004000230 2006 eng d00aUsing Query-Speci_c Variance Estimates to Combine Bayesian Classi_ers0 aUsing QuerySpecic Variance Estimates to Combine Bayesian Classie1 aGreiner, R.1 aWang, Shaojun1 aLee, C. uhttp://knoesis.wright.edu/node/106501728nas a2200133 4500008004100000245008300041210006900124520129800193100001301491700001301504700001801517700001901535856004001554 2005 eng d00aCombining Statistical Language Models via the Latent Maximum Entropy Principle0 aCombining Statistical Language Models via the Latent Maximum Ent3 aWe present a unified probabilistic framework for statistical language modeling which can simultaneously incorporate various aspects of natural language, such as local word interaction, syntactic structure and semantic document information. Our approach is based on a recent statistical inference principle we have proposed-the latent maximum entropy principle-which allows relationships over hidden features to be effectively captured in a unifiedmodel. Our work extends previous research on maximum entropy methods for language modeling, which only allow observed features to be modeled. The ability to conveniently incorporate hidden variables allows us to extend the expressiveness of language models while alleviating the necessity of pre-processing the data to obtain explicitly observed features.We describe efficient algorithms for marginalization, inference and normalization in our extended models. We then use these techniques to combine two standard forms of language models: local lexical models (Markov N-gram models) and global document-level semantic models (probabilistic latent semantic analysis). Our experimental results on the Wall Street Journal corpus show that we obtain a 18.5% reduction in perplexity compared to the baseline tri-gram model with Good-Turing smoothing.1 aPeng, F.1 aZhao, Y.1 aWang, Shaojun1 aSchuurmans, D. uhttp://knoesis.wright.edu/node/145600484nas a2200133 4500008004100000245011500041210006900156100001400225700001900239700001600258700001800274700001800292856004000310 2005 eng d00aExploiting Syntactic, Semantic and Lexical Regularities in Language Modeling via Directed Markov Random Fields0 aExploiting Syntactic Semantic and Lexical Regularities in Langua1 aCheng, L.1 aSchuurmans, D.1 aGreiner, R.1 aWang, Shaojun1 aWang, Shaojun uhttp://knoesis.wright.edu/node/107400349nas a2200121 4500008004100000245004100041210004100082100001300123700001800136700001900154700001400173856004000187 2005 eng d00aVariational Bayesian Image Modelling0 aVariational Bayesian Image Modelling1 aJiao, F.1 aWang, Shaojun1 aSchuurmans, D.1 aCheng, L. uhttp://knoesis.wright.edu/node/114600395nas a2200109 4500008004100000245008500041210006900126100001300195700001800208700001900226856004000245 2004 eng d00aAugmenting Naive Bayes Text Classi_er Using Statistical N-Gram Language Modeling0 aAugmenting Naive Bayes Text Classier Using Statistical NGram Lan1 aPeng, F.1 aWang, Shaojun1 aSchuurmans, D. uhttp://knoesis.wright.edu/node/146100528nas a2200145 4500008004100000245011500041210006900156260002500225100001800250700001800268700002100286700002100307700001400328856004000342 2004 eng d00aExploiting Syntactic, Semantic and Lexical Regularities in Language Modeling via Directed Markov Random Fields0 aExploiting Syntactic Semantic and Lexical Regularities in Langua aSingapore, Singapore1 aWang, Shaojun1 aWang, Shaomin1 aGreiner, Russell1 aSchuurmans, Dale1 aCheng, Li uhttp://knoesis.wright.edu/node/107301057nas a2200133 4500008004100000245008200041210006900123520062800192100001900820700001300839700001300852700001800865856004000883 2004 eng d00aLearning Mixture Models with the Regularized Latent Maximum Entropy Principle0 aLearning Mixture Models with the Regularized Latent Maximum Entr3 aWe present a new approach to estimating mixture models based on a new inference principle we have proposed: the latent maximum entropy principle (LME). LME is different both from Jaynes' maximum entropy principle and from standard maximum likelihood estimation. We demonstrate the LME principle by deriving new algorithms for mixture model estimation, and show how robust new variants of the EM algorithm can be developed. Our experiments show that estimation based on LME generally yields better results than maximum likelihood estimation, particularly when inferring latent variable models from small amounts of data.1 aSchuurmans, D.1 aPeng, F.1 aZhao, Y.1 aWang, Shaojun uhttp://knoesis.wright.edu/node/145800408nas a2200121 4500008004100000245007300041210006900114100001300183700001800196700001900214700001300233856004000246 2003 eng d00aBoltzmann Machine Learning with the Latent Maximum Entropy Principle0 aBoltzmann Machine Learning with the Latent Maximum Entropy Princ1 aZhao, Y.1 aWang, Shaojun1 aSchuurmans, D.1 aPeng, F. uhttp://knoesis.wright.edu/node/106800409nas a2200109 4500008004100000245009900041210006900140100001800209700001300227700001900240856004000259 2003 eng d00aLanguage and Task Independent Text Categorization Using Character Level N-Gram Language Models0 aLanguage and Task Independent Text Categorization Using Characte1 aWang, Shaojun1 aPeng, F.1 aSchuurmans, D. uhttp://knoesis.wright.edu/node/155400414nas a2200109 4500008004100000245010400041210006900145100001300214700001800227700001900245856004000264 2003 eng d00aLanguage Independent Automated Authorship Attribution with Character Level N-Gram Language Modeling0 aLanguage Independent Automated Authorship Attribution with Chara1 aPeng, F.1 aWang, Shaojun1 aSchuurmans, D. uhttp://knoesis.wright.edu/node/113901142nas a2200121 4500008004100000245007400041210006900115520074600184100001300930700001800943700001900961856004000980 2003 eng d00aLatent Maximum Entropy Approach for Semantic N-gram Language Modeling0 aLatent Maximum Entropy Approach for Semantic Ngram Language Mode3 aIn this paper, we describe a unified probabilistic framework for statistical language modeling--the latent maximum entropy principle--which can effectively incorporate various aspects of natural language, such as local word interaction, syntactic structure and semantic document information. Unlike previous work on maximum entropy methods for language modeling, which only allow explicit features to be modeled, our framework also allows relationships over hidden features to be captured, resulting in a more expressive language model. We describe efficient algorithms for marginalization, inference and normalization in our extended models. We then present promising experimental results for our approach on the Wall Street Journal corpus.1 aPeng, F.1 aWang, Shaojun1 aSchuurmans, D. uhttp://knoesis.wright.edu/node/182500357nas a2200097 4500008004100000245007200041210006900113100001800182700001900200856004000219 2003 eng d00aLearning Continuous Latent Variable Models with Bregman Divergences0 aLearning Continuous Latent Variable Models with Bregman Divergen1 aWang, Shaojun1 aSchuurmans, D. uhttp://knoesis.wright.edu/node/106700338nas a2200097 4500008004100000245006100041210006100102100001800163700001900181856004000200 2003 eng d00aLearning Latent Variable Models with Bregman Divergences0 aLearning Latent Variable Models with Bregman Divergences1 aWang, Shaojun1 aSchuurmans, D. uhttp://knoesis.wright.edu/node/107000405nas a2200121 4500008004100000245007000041210006900111100001300180700001800193700001300211700001900224856004000243 2003 eng d00aLearning Mixture Models with the Latent Maximum Entropy Principle0 aLearning Mixture Models with the Latent Maximum Entropy Principl1 aZhao, Y.1 aWang, Shaojun1 aPeng, F.1 aSchuurmans, D. uhttp://knoesis.wright.edu/node/106900415nas a2200121 4500008004100000245008000041210006900121100001900190700001300209700001300222700001800235856004000253 2003 eng d00aSemantic N-gram Language Modeling with the Latent Maximum Entropy Principle0 aSemantic Ngram Language Modeling with the Latent Maximum Entropy1 aSchuurmans, D.1 aPeng, F.1 aZhao, Y.1 aWang, Shaojun uhttp://knoesis.wright.edu/node/114300405nas a2200121 4500008004100000245006900041210006900110100001900179700001800198700001300216700001400229856004000243 2003 eng d00aText Classification in Asian Languages Without Word Segmentation0 aText Classification in Asian Languages Without Word Segmentation1 aSchuurmans, D.1 aWang, Shaojun1 aPeng, F.1 aHuang, X. uhttp://knoesis.wright.edu/node/114500349nas a2200121 4500008004100000245004100041210003700082100001800119700001300137700001900150700001800169856004000187 2002 eng d00aThe Latent Maximum Entropy Principle0 aLatent Maximum Entropy Principle1 aWang, Shaojun1 aZhao, Y.1 aSchuurmans, D.1 aRosenfeld, R. uhttp://knoesis.wright.edu/node/107100362nas a2200133 4500008004100000245003600041210003600077100001500113700001800128700001300146700001500159700001400174856004000188 2002 eng d00aPredicting Oral Reading Miscues0 aPredicting Oral Reading Miscues1 aWinter, V.1 aWang, Shaojun1 aBeck, J.1 aMostow, J.1 aTobin, B. uhttp://knoesis.wright.edu/node/114000880nas a2200109 4500008004100000245009600041210006900137520049300206100001300699700001800712856004000730 2001 eng d00aAlmost Sure Convergence of Titterington's Recursive Estimator for Finite Mixture Models0 aAlmost Sure Convergence of Titterington39s Recursive Estimator f3 aTitterington proposed a recursive parameter estimation algorithm for finite mixture models. However, due to the well known problem of singularities and multiple maximum, minimum and saddle points that are possible on the likelihood surfaces, convergence analysis has seldom been made in the past years. In this paper, under mild conditions, we show the global convergence of Titterington's recursive estimator and its MAP variant for mixture models of full regular exponential family.1 aZhao, Y.1 aWang, Shaojun uhttp://knoesis.wright.edu/node/114200371nas a2200097 4500008004100000245009200041210006900133100001800202700001300220856004000233 2001 eng d00aAlmost Sure Convergence of Titterington's Recursive Estimator for Finite Mixture Models0 aAlmost Sure Convergence of Titteringtons Recursive Estimator for1 aWang, Shaojun1 aZhao, Y. uhttp://knoesis.wright.edu/node/235700380nas a2200109 4500008004100000245007100041210006900112100001800181700001300199700001800212856004000230 2001 eng d00aLatent Maximum Entropy Principle for Statistical Language Modeling0 aLatent Maximum Entropy Principle for Statistical Language Modeli1 aWang, Shaojun1 aZhao, Y.1 aRosenfeld, R. uhttp://knoesis.wright.edu/node/182601623nas a2200109 4500008004100000245011200041210006900153520122000222100001301442700001801455856004001473 2001 eng d00aOn-Line Bayesian Tree-Structured Transformation of HMMs with Optimal Model Selection for Speaker Adaptation0 aOnLine Bayesian TreeStructured Transformation of HMMs with Optim3 aThis paper presents a new recursive Bayesian learning approach for transformation parameter estimation in speaker adaptation. Our goal is to incrementally transform or adapt a set of hidden Markov model (HMM) parameters for a new speaker and gain large performance improvement from a small amount of adaptation data. By constructing a clustering tree of HMM Gaussian mixture components, the linear regression (LR) or affine transformation parameters for HMM Gaussian mixture components are dynamically searched. An online Bayesian learning technique is proposed for recursive maximum a posteriori (MAP) estimation of LR and affine transformation parameters. This technique has the advantages of being able to accommodate flexible forms of transformation functions as well as a priori probability density functions (pdfs). To balance between model complexity and goodness of fit to adaptation data, a dynamic programming algorithm is developed for selecting models using a Bayesian variant of the 'minimum description length' (MDL) principle. Speaker adaptation experiments with a 26-letter English alphabet vocabulary were conducted, and the results confirmed effectiveness of the online learning framework.1 aZhao, Y.1 aWang, Shaojun uhttp://knoesis.wright.edu/node/146200387nas a2200109 4500008004100000245008400041210006900125100001200194700001800206700001300224856004000237 2001 eng d00aRecursive Estimation of Time-Varying Environments for Robust Speech Recognition0 aRecursive Estimation of TimeVarying Environments for Robust Spee1 aYen, K.1 aWang, Shaojun1 aZhao, Y. uhttp://knoesis.wright.edu/node/114100377nas a2200097 4500008004100000245009800041210006900139100001800208700001300226856004000239 2000 eng d00aOn-Line Bayesian Speaker Adaptation By Using Tree-Structured Transformation and Robust Priors0 aOnLine Bayesian Speaker Adaptation By Using TreeStructured Trans1 aWang, Shaojun1 aZhao, Y. uhttp://knoesis.wright.edu/node/115600345nas a2200097 4500008004100000245006800041210006700109100001800176700001300194856004000207 2000 eng d00aOptimal On-Line Bayesian Model Selection for Speaker Adaptation0 aOptimal OnLine Bayesian Model Selection for Speaker Adaptation1 aWang, Shaojun1 aZhao, Y. uhttp://knoesis.wright.edu/node/115700378nas a2200097 4500008004100000245009900041210006900140100001800209700001300227856004000240 1999 eng d00aOn-Line Bayesian Tree-Structured Transformation of Hidden Markov Models for Speaker Adaptation0 aOnLine Bayesian TreeStructured Transformation of Hidden Markov M1 aWang, Shaojun1 aZhao, Y. uhttp://knoesis.wright.edu/node/115500370nas a2200097 4500008004100000245009100041210006900132100001800201700001300219856004000232 1999 eng d00aA Unifed Framework for Recursive Maximum Likelihood Estimation of Hidden Markov Models0 aUnifed Framework for Recursive Maximum Likelihood Estimation of 1 aWang, Shaojun1 aZhao, Y. uhttp://knoesis.wright.edu/node/115400385nas a2200109 4500008004100000245008300041210006900124100001800193700001100211700001300222856004000235 1998 eng d00aOn Convergence of Maximum Likelihood Estimation of Binary HMMs by EM Algorithm0 aConvergence of Maximum Likelihood Estimation of Binary HMMs by E1 aWang, Shaojun1 aLi, M.1 aZhao, Y. uhttp://knoesis.wright.edu/node/115900406nas a2200109 4500008004100000245010200041210006900143100001800212700001200230700001400242856004000256 1997 eng d00aProbabilistic Production Costing of Hydro and Pumped Storage Units under Chronological Load Curve0 aProbabilistic Production Costing of Hydro and Pumped Storage Uni1 aWang, Shaojun1 aXia, Q.1 aXiang, N. uhttp://knoesis.wright.edu/node/146500362nas a2200109 4500008004100000245005900041210005900100100002100159700001800180700001400198856004000212 1995 eng d00aProbabilistic Marginal Cost Curve and Its Applications0 aProbabilistic Marginal Cost Curve and Its Applications1 aShahidehpour, S.1 aWang, Shaojun1 aXiang, N. uhttp://knoesis.wright.edu/node/145900500nas a2200133 4500008004100000245012600041210006900167100002100236700001800257700001700275700001700292700001700309856004000326 1995 eng d00aShort-Term Generation Scheduling with Transmission and Environmental Constraints Using an Augmented Lagrangian Relaxation0 aShortTerm Generation Scheduling with Transmission and Environmen1 aShahidehpour, S.1 aWang, Shaojun1 aMokhtari, S.1 aKirschen, D.1 aIrisarri, G. uhttp://knoesis.wright.edu/node/146000371nas a2200109 4500008004100000245006800041210006800109100001400177700001200191700001800203856004000221 1994 eng d00aProbabilistic Production Costing under Chronological Load Curve0 aProbabilistic Production Costing under Chronological Load Curve1 aXiang, N.1 aXia, Q.1 aWang, Shaojun uhttp://knoesis.wright.edu/node/1473