Mutual Information & Dialog Systems

Mutual Information

See also:

WLMI | Weighted Mutual Information

Data selection for language modeling using sparse representations A Sethy, TN Sainath, B Ramabhadran… – … Annual Conference of …, 2010 – … a vocabulary of 2K words from the test set which had high mutual information with the … The weight vector obtained by solving this lin- ear regression problem is then used to … Weilhammer, MN Stuttlem, and S. Young, “Boot- strapping language models for dialogue systems,” in Pro … Cited by 1 – Related articles – All 3 versions

F 2–new technique for recognition of user emotional states in spoken dialogue systems [PDF] from R López-Cózar, J Silovsky… – … of the 11th Annual Meeting of the …, 2010 – … For example, these have been applied to spoken dialogue systems (SDSs) used in automated call-centres … ie, a weighted linear combination of Q uni- modal Gaussian densities Pl(x). The … word for a given emotional category can be defined as the mutual information between the … Cited by 1 – Related articles – All 8 versions

[PDF] Concept Type Prediction and Responsive Adaptation in a Dialogue System [PDF] from S Stoyanchev, AJ Stent – Dialogue and Discourse, 2012 – … By using a weighted combination of topic-specific language models, they obtained … and shorter sequences of speech recognition errors, and improved dialogue system performance … the most salient lexical features: manual identification and mutual information-based identification … Related articles – View as HTML – All 2 versions

Domain adaptation with unlabeled data for dialog act tagging [PDF] from A Margolis, K Livescu… – … of the 2010 Workshop on Domain …, 2010 – … applied supervised model adaptation to intent classification across customer dialog systems, and Guz … that they be common in both domains and have high mutual information (MI) with … an improvement in Backchannel recall, which occurred under both weighted and unweighted … Cited by 6 – Related articles – All 10 versions

AAM based continuous facial expression recognition for face image sequences [PDF] from S Hommel… – Computational Intelligence and …, 2011 – … However, emotion recognition is needed to generate a natural dialog system. … Furthermore, the mutual information between all parameters became calcu- lated … correlation values cv of every dimension i with the positivity values is computed to obtain the weighted parameters Cpi … Related articles – All 2 versions

Multi-state barge-in models for spoken dialog systems A Ljolje – US Patent 8,046,221, 2011 – Google Patents … 704/246 SPOKEN DIALOG SYSTEMS 2°03/0088411 A1* 5/2°03 M=1ete1_~ ~~~~ 704/236 2005/0027527 A1* 2/2005 Junkawitsch et al. …. … The principles of this system may be utilized to facilitate a user’s interaction which spoken dialogue systems. … Related articles – All 4 versions

Unified treatment of data-sparseness and data-overfitting in maximum entropy modeling F Weng… – EP Patent 1,783,744, 2011 – … are represented with features and each feature is assigned a weight based on … For the second component, the mutual information component is further approximated through maximum … in above section entitled “Modeling Language Understanding for Dialog Systems”, in which … Related articles – Cached

Dialogue management based on entities and constraints [PDF] from Y Xu… – Proceedings of the 11th Annual Meeting of the Special …, 2010 – … 2 Related Work In recent years, statistical methods have gained popularity in dialogue system research. … The weight for each dimension wi is proportional to the count of distinct values of the particular dimen- sion c(Di) and the mutual information between the dimension … Cited by 3 – Related articles – All 14 versions

[PDF] Unsupervised Classification of Biomedical Abstracts using Lexical Association? [PDF] from J Read, J Webster… – 2010 – … Keywords: Pointwise mutual information, Text classification, Unsupervised methods … The RAMCORP project aims to design and construct a telephony-based dialogue system that pro- vides … We weighted descriptors marked as major topics (or those with a major topic qualifier) as … Related articles – View as HTML – All 5 versions

MULTI-STATE BARGE-IN MODELS FOR SPOKEN DIALOG SYSTEMS A Ljolje – US Patent 20,120,101,820, 2012 – … models, this achieved by changing the network (language model) cost weight when doing … barge-in acoustic model may further be trained using maximum mutual information (MMI) criterion … in model that is later used for speech processing preferably in a spoken dialog system. … Cached

Impact of word classing on shrinkage-based language models [PDF] from R Sarikaya, SF Chen, A Sethy… – … Annual Conference of …, 2010 – … this using the IBM Infinite State Machine Toolkit [6]. By using the weighted automata framework, it … Pairs of classes with the least average loss in mutual information loss are merged. … Predefined Classes Most of the spoken dialog systems have domain specific vocab- ulary, which … Cited by 2 – Related articles – All 3 versions

Index-based incremental language model for scalable directory assistance A Moreno-Daniel, J Wilpon… – Speech Communication, 2011 – Elsevier … 1. Introduction Voice search enables an informational spoken dialog system (SDS) to find a ranked … name- grammar followed by a last-name-grammar, therefore discarding any inter-field mutual information. … let E ∈ (E 1 ,…,E j ,…,EN ) represent the corresponding weighted set of … Related articles – All 2 versions

[CITATION] Acoustic modeling for robust utterance verification in continuous speech… [PDF] from BH Juang… – 2011 – … The main application of this year’s project is a spoken dialog system for English education … Several Discriminative Training (DT) methods, such as maximum mutual information estimation (MMIE) [3], minimum … where ? is the HMM parameter set and ( , |?) is a weighted sum over … Related articles

Identifying problematic dialogs in a human-computer dialog system [PDF] from HC Truong – 2010 – … al., 1999) which addressed the automatic prediction of problematic Human-Computer dialogues with their system ‘How May I Help You?’ HMIHY is a spoken dialogue system for customer care … length, age, weight and so on). • Class … Related articles – All 2 versions

[PDF] On-Line Linear Combination of Classifiers Based on Incremental Information in Speaker Verification [PDF] from F Huenupán, NB Yoma, C Garretón… – ETRI Journal, 2010 – … 2(b)). In this paper, the mutual information criterion is applied to address the problem of on-line optimization of multiclassifier fusion in text-dependent (TD) SV with … The addressed MCS scheme is the weighted linear combination (also known as the sum rule) of classifier scores. … Related articles – View as HTML – All 2 versions

Discriminative training of multi-state barge-in models for speech processing A Ljolje – US Patent 8,000,971, 2011 – Google Patents … No.: 11/930,656 (57) ABSTRACT _ Disclosed are systems and methods for training a barge-in- (22) Flledi 0et- 31: 2007 model for speech processing in a spoken dialogue system comprising the steps of (1) receiving an input having at least (65) Pl’i0l’ PllbII¢3ti0l1 Data one … Related articles – All 5 versions

Speech emotion recognition using FCBF feature selection method and GA-optimized fuzzy ARTMAP neural network D Gharavian, M Sheikhan, A Nazerieh… – Neural Computing & …, 2011 – Springer … The filter method employs intrinsic properties of data, such as mutual information, as the criterion for feature subset evaluation. … It is noted that mutual information (MI) of two vectors X and Y, I(X,Y), computes the sta- tistical dependency of them in the following way: IðX,YÞ 1/4 ? … Cited by 5 – Related articles

Multimodal fusion for multimedia analysis: a survey [PDF] from PK Atrey, MA Hossain, A El Saddik… – Multimedia Systems, 2010 – Springer … The mutual information is a measure of the information of one modality conveyed about another … As can be seen from the table, in rule-based fusion category, linear weighted fusion method … This fusion method is widely used in the domain of multimodal dialog systems and sports … Cited by 29 – Related articles – All 9 versions

Anger recognition in speech using acoustic and linguistic cues [PDF] from T Polzehl, A Schmitt, F Metze… – Speech Communication, 2011 – Elsevier … b Dialogue Systems Group / Institute of Information Technology, University of Ulm, Albert-Einstein-Allee 43 … Statistical in- dependence of the two probabilities would result in a mutual information of zero … the graphs of the unweighted aver- age recall and the weighted average recall … Cited by 4 – Related articles – All 6 versions

Salient Features for Anger Recognition in German and English IVR Portals [PDF] from T Polzehl, A Schmitt… – Spoken Dialogue Systems Technology …, 2011 – Springer … Emotion detection in Interactive Voice Response (IVR) dialogue systems can be used to monitor … the class-dependent discriminative power of a word using the self-mutual information criterion … class-independent score of emotional salience he applies a weighted summation over … Cited by 2 – Related articles – All 2 versions

[PDF] Learning from Imbalanced and Incomplete Data: a Practical Approach [PDF] from W Chen – 2011 – … The main focus of this thesis is to automatically detect off-task speech in children’s tutorial dialogs. Off-task speech poses different levels of technical challenges to children’s tutorial dialog systems. … 2. for 3. do train weak classifiers on weighted 4. choose … Related articles – View as HTML

Speech and multimodal interaction in mobile search [PDF] from J Feng, M Johnston… – Signal Processing …, 2011 – … Reference [19] segmented a query based on the mutual information (MI) between pairs of query words. … The arcs are weighted by c1wd,d2, which is based on tf.idf score. … While telephony-based unimodal dialog systems lean toward a more structured interaction where the user … Cited by 2 – Related articles – All 5 versions

Benchmarking short text semantic similarity J O’Shea, Z Bandar, K Crockett… – International Journal of …, 2010 – Inderscience … Lapalme, 2004), information retrieval (IR) (Marton and Katz, 2006), healthcare dialogue systems (Bickmore and … Kennedy and Szpakowicz 2008 Wordnet (weighted) plus cosine measure … Inkpen (2008) combine a variant of pointwise mutual information (using the British National … Cited by 1 – Related articles – All 6 versions

Intent Determination and Spoken Utterance Classification G Tur, L Deng – Spoken Language Understanding, 2011 – Wiley Online Library … the relation between the class C and word sequence W. Typically, binary or weighted n-gram … Should the prompts/actions taken by the dialogue system change over time the semantic representation must also … where ck is a call-type and I(x, y) is the mutual information between x … Related articles – All 3 versions

Emotion recognition in text for 3-D facial expression rendering RA Calix, SA Mallepudi, B Chen… – … , IEEE Transactions on, 2010 – … These interfaces can serve as dialogue systems which use speech and facial expressions to … After trying several weight combinations, a set was identified which consistently provided better … 1 presents the results of the classification task using mutual information feature selection … Cited by 8 – Related articles – All 4 versions

Mining association language patterns using a distributional semantic model for negative life event classification LC Yu, CL Chan, CC Lin… – Journal of Biomedical Informatics, 2011 – Elsevier … A dialog system could generate supportive responses if it could understand the negative life events embedded in … patterns, all frequent word sets were sorted in descending order of their mutual information scores. … kN wwcwc wc vdd d= (2) where ki wc d denotes the weight of the … Related articles – All 3 versions

Document sentiment classification by exploring description model of topical terms Y Hu… – Computer Speech & Language, 2011 – Elsevier … 2011, Pages 386-403 Language and speech issues in the engineering of companionable dialogue systems. … In constructing the MST structure, the weight of the link between a pair (¡satisfy, audience¿ for example) is measured by the Pointwise Mutual Information (PMI) (eg … Cited by 5 – Related articles – All 3 versions

Unsupervised semantic similarity computation between terms using Web documents [PDF] from E Iosif… – Knowledge and Data Engineering, …, 2010 – … language modeling [5], grammar induction [6], word sense disambiguation [7], and speech understanding and spoken dialogue systems [5]. In … 1/4 Y , the knowledge of X provides the value of Y with certainty and the mutual information is 1 … The binary metric assigns weight tw;i 1/4 1 … Cited by 13 – Related articles – All 12 versions

CoCITe-Coordinating Changes in Text [HTML] from J Wright… – Knowledge and Data Engineering, …, 2012 – … Unfortunately, this maximization requires a nonlinear iteration which compromises the speed of the procedure in an unacceptable way. Instead, the parameters are estimated using weighted linear regression. … 8 >>< >> The total weighted squared error for the ith segment is then … Related articles – All 18 versions

Margin-based discriminative training for string recognition G Heigold, P Dreuw, S Hahn… – Selected Topics in …, 2010 – … A. Maximum Mutual Information (MMI) The MMI training criterion is defined as (3) … Lemma 1 (M-MMI/Hinge): . M-MPE in (9) implies a (weighted) margin error (eg, phoneme error) combined with a weighted margin (17) Page 6. 922 … Cited by 7 – Related articles – All 3 versions

Annotating and identifying emotions in text [PDF] from C Strapparava… – Intelligent Information Access, 2010 – Springer … The emotion scores were obtained through Pointwise Mutual Information (PMI). … Second, associative score between a content word and an emotion was estimated and used to weight the final PMI score. The obtained results were normalized in the range 0-100. SWAT: … Cited by 3 – Related articles – All 3 versions

Discriminative training of HMMs for automatic speech recognition: A survey [PDF] from H Jiang – Computer Speech & Language, 2010 – Elsevier … Maximum mutual information estimation (MMIE); 2.2. … has been significantly improved in a variety of real-world applications, from simple digit recognition to large vocabulary broadcast news transcription, from reading style voice dictation to spontaneous dialogue systems, etc. … Cited by 19 – Related articles – All 6 versions

Improving Spontaneous Children’s Emotion Recognition by Acoustic Feature Selection and Feature-Level Fusion of Acoustic and Linguistic Parameters S Planet… – Advances in Nonlinear Speech Processing, 2011 – Springer … the automatic generation of audiovisual content, for virtual meetings or even in automatic dialogue systems. … are the use of only one learning stage and taking advantage of mutual information. … 1). However, in most of the studies of emotion recognition the weighted average recall … Related articles – All 3 versions

Maximum entropy-based reinforcement learning using a confidence measure in speech recognition for telephone speech C Molina, NB Yoma, F Huenupán… – Audio, Speech, and …, 2010 – … comparable with discriminative training where models are trained. Finally, the results presented should be important to some commercial applications such as telephone dialogue systems. 2 Bayes-based confidence measure … Cited by 3 – Related articles – All 4 versions

Bayesian Networks for Discrete Observation Distributions in Speech Recognition A Miguel, A Ortega, L Buera… – Audio, Speech, and …, 2011 – … the covariance ma- trix structure [13], [14], for language modeling [15] or in dialog systems [16]. … and store in memory all the joint count statistics needed to calculate the mutual information terms for … A mixture of C components is usually defined as a weighted sum of C specific … Cited by 2 – Related articles – All 2 versions

[PDF] Introduction to Discriminative Training in Speech Recognition [PDF] from R Schlüter… – 2010 – … standard maximum likelihood (ML) training: maximize reference class conditional p?(x|c) ? maximum mutual information (MMI) training: maximize … covers maximum mutual information (MMI), minimum classification error (MCE), minimum phone/word error (MPE/MWE) … Related articles – View as HTML

[PDF] Bayesian Approaches to Uncertainty in Speech Processing [PDF] from PN Garner – 2011 – … However, his main contribution was perhaps more subtle: At the time, it was popular to use a technique known as weight decay to penalise large weights in the MLP. MacKay showed that weights could be penalised by means of a (Gaussian) prior. … View as HTML

Model Shrinkage for Discriminative Language Models OBA Takanobu, H Takaaki… – … on Information and …, 2012 – … consist of I utterances. • Word error rate (WER) as sample weight The WER of each word sequence hypothesis is used for sample weight training, where a sample is a word sequence in the lists. The parameters are estimated …

A large vocabulary continuous speech recognition system for Persian language [PDF] from H Sameti, H Veisi, M Bahrani, B Babaali… – EURASIP Journal on …, 2011 – Springer … In the last decade, researchers and investors introduced spoken dialogue systems and tried to implement … these clustering methods, certain information theory cri- teria, such as average mutual information, are used … wn+1) and SLM(wn+1) are differ- ent, a weight parameter (aLM … Related articles – All 3 versions

Learning to recognize objects using waves of spikes and Spike Timing-Dependent Plasticity [PDF] from T Masquelier… – Neural Networks (IJCNN), The …, 2010 – … The winner thus fires and the STDP rule is triggered. The weight matrix is updated, and the change in weights is duplicated at all positions and scales. This … al. proposed an interesting criterion based on mutual information [46]. … Cited by 5 – Related articles – All 5 versions

Large Margin Estimation of Hidden Markov Models With Second Order Cone Programming for Speech Recognition [PDF] from D Wu, Y Yin… – Audio, Speech, and Language …, 2011 – … mixture hidden Markov models (CDHMMs) and a variety of DT methods have been investigated for CDHMMs, such as maximum mutual information estimation (MMIE … where pk and wk are initial probability and weight for the k-th Gaussian and skd is the d-dimensional variance … Cited by 1 – Related articles – All 3 versions


Efficient Blind Dereverberation and Echo Cancellation Based on Independent Component Analysis for Actual Acoustic Signals R Takeda, K Nakadai, T Takahashi, K Komatani… – Neural …, 2011 – MIT Press … This method enables spoken dialogue systems to handle barge-in situations (Komatani, Kawahara, & Okuno, 2008; Matsuyama, Komatani, Ogata, & Okuno, 2009). … The result of ASR is used in the spoken dialogue system to produce system utterances. … Related articles – All 3 versions

Cross-language information retrieval [PDF] from JY Nie – Synthesis Lectures on Human Language …, 2010 – … Semantic Role Labeling Martha Palmer, Daniel Gildea, Nianwen Xue 2010 Spoken Dialogue Systems Kristiina Jokinen, Michael McTear 2010 … Term weighting has been integrated into extended Boolean models so that a document is represented by a set of weighted terms. … Cited by 17 – Related articles – Library Search – All 15 versions

Speech Recognition by Machine, A Review [PDF] from MA Anusuya… – Arxiv preprint arXiv:1001.2267, 2010 – Page 1. (IJCSIS) International Journal of Computer Science and Information Security, Vol. 6, No. 3, 2009 Speech Recognition by Machine: A Review MAAnusuya Department of Computer Science and Engineering Sri Jaya chamarajendra College of Engineering Mysore, India . … Cited by 24 – Related articles – All 7 versions

Recognising realistic emotions and affect in speech: State of the art and lessons learnt from the first challenge [PDF] from B Schuller, A Batliner, S Steidl… – Speech Communication, 2011 – Elsevier … As an alternative, a few studies modelled low-level features directly, mainly by Gaussian Mixture Models (GMM, ie, continuous HMM with a single state that use weighted mixtures of Gaussian densities) [135, 209] and HMM [137, 141, 188, 247, 94, 139, 231]. … Cited by 36 – Related articles – All 9 versions

Medical data mining: Improving information accessibility using online patient drug reviews [PDF] from S Seneff, YA Li – 2011 – … 25 2.2 Medical Applications….. 26 2.2.1 Dialogue Systems….. 26 2.2.2 Health Surveillance….. … 4.3.1 Log Likelihood Statistic….. 44 4.3.2 Pointwise Mutual Information…. . 45 4.3.3 Set Operations. …. … Cited by 2 – Related articles – All 3 versions

Discriminative training and acoustic modeling for automatic speech recognition [PDF] from W Macherey – 2010 – … 16 2.7 Maximum Mutual Information estimates for a two-class problem. … forms an integral part in many applications such as spoken document retrieval [Macherey & Viechtbauer` 03], speech-to-speech translation [Matusov & Kanthak` 05], or spoken dialog systems [Macherey … Cited by 1 – Related articles – Library Search – All 10 versions

Paralinguistics in speech and language-State-of-the-art and the challenge B Schuller, S Steidl, A Batliner, F Burkhardt… – Computer Speech & …, 2012 – Elsevier … For instance, we assume only two genders even if there exist more varieties in between. • Long term traits: -. biological trait primitives such as height ( [Mporas and Ganchev, 2009] and [Schuller et al., 2011e]), weight, age (Schuller et al., 2010b), gender (Schuller et al., 2010b); – … Cited by 1

[HTML] Frontiers: The Timing of Vision-How Neural Processing Links to Different Temporal Dynamics [HTML] from T Masquelier, L Albantakis… – Frontiers in Perception …, 2011 – Home; About; Submit; Advertise & PR; Alerts. Register; Login. Science: Genetics: Applied Genetic Epidemiology; Behavioral and Psychiatric Genetics; Bioinformatics and Computational Biology; Cancer Genetics; Epigenomics; Evolutionary … Related articles – Cached – All 4 versions

[PDF] Computational approaches for emotion detection in text [PDF] from H Binali, C Wu… – Digital Ecosystems and …, 2010 – … wordNet1.6. Emotional weight, OMCS (Open mind common sense) knowledge base … Using Bio-sensors: First Steps towards an Automatic System ” in Affective Dialogue Systems. vol. … [14] J. Read, “Recognising Affect in Text using Pointwise- Mutual Information “: Master of … Cited by 2 – Related articles – View as HTML – All 2 versions

[PDF] Review of Korean Speech Act Classification: Machine Learning Methods [PDF] from H Kim, CN Seon… – Journal of Computing Science and …, 2011 – … If a dialogue system fails to capture users’ intentions, the system will not be able to decide … The goal of MLP is to find the set of weight values that will cause the … They showed that the ? 2 statistic outperforms mutual information and information gain in document classification. … Related articles – View as HTML – All 2 versions

[PDF] Recognising Emotions and Sentiments in Text [PDF] from S Mac Kim – 2011 – … Emotion Antecedents and Reactions ITSs Intelligent Tutoring Systems ITSPOKE Intelligent Tutoring SPOKEn dialogue system KBANN Knowledge … Sense PAD Pleasure-Arousal-Dominance PLSA Probabilistic Latent Semantic Analysis PMI Pointwise Mutual Information SAM Self … View as HTML

[HTML] The timing of vision-how neural processing links to different temporal dynamics [HTML] from T Masquelier, L Albantakis… – Frontiers in psychology, 2011 – … The biased-competition theory claims that the neuronal response – in terms of firing rate – to simultaneously presented stimuli is a weighted average of the response to isolated stimuli, and that attention biases the weights in favor of the attended stimulus (Desimone. Cited by 2 – Related articles – All 6 versions

Which words are hard to recognize? Prosodic, lexical, and disfluency factors that increase speech recognition error rates [PDF] from S Goldwater, D Jurafsky… – Speech Communication, 2010 – Elsevier … et al.’s (2004) analysis of two human-computer dialogue systems, misrecognized turns … factors combined into a single predictor known as frequency-weighted neighborhood density … variants of minimum phone error (MPE) training with maximum mutual information (MMI) priors … Cited by 20 – Related articles – All 36 versions

[BOOK] Contextual computing: Models and applications R Porzel – 2010 – … Chapter4 discusses integratinguser andsituation knowledgeinto dialog systems. … Such pragmatic patterns can then be used together with context data in order to help determine how a dialog system should react when multiple solutions are possible. … Cited by 1 – Related articles – Library Search – All 3 versions

Context-Dependent Pre-trained Deep Neural Networks for Large Vocabulary Speech Recognition [PDF] from G Dahl, D Yu, L Deng… – Audio, Speech, and …, 2010 – … eg, [2], [3]). There have been some notable recent advances in discrimina- tive training (see an overview in [4]; eg, maximum mutual information (MMI) estimation … In the experiments in this work, we use the following expression for the t + 1st weight update for some typical model … Cited by 11 – Related articles – All 7 versions

Equivalence of Generative and Log-Linear Models G Heigold, H Ney, P Lehnen… – Audio, Speech, and …, 2011 – … Assuming a weighted finite-state transducer (WFST) with non-negative arc weights, weight pushing produces an equivalent stochastic WFST [13, p.242]. … 3 criterion. The conventional choice of training criterion is the maximum mutual information (MMI)1 F(?) = N ? … Cited by 1 – Related articles – All 2 versions

[BOOK] Speech Processing and Soft Computing SA Selouani – 2011 – … Numerous spectral domain measures have been proposed in the literature including the log-likelihood ratio (LLR) measures [27], the cepstral distance measures [134], and the weighted slope spectral distance measure (WSS) [79]. … Related articles

[PDF] Deliverable 5.1 [PDF] from JM RBB, J Bán, M Beniak, M Féder, J Kacur… – 2012 – … Page 11 Approximation of the probability density distribution by a weighted sample set St = { , | i = 1, · · ·,Np} (2.1) Each sample represents a hypothetical state of the object, and … to a system model. Each sample element in the set is weighted in terms of the observations, … View as HTML

WordICA-emergence of linguistic representations for words by independent component analysis [PDF] from T Honkela, A HyvÄrinen… – Natural Language …, 2010 – Cambridge Univ Press … Dialogue systems can benefit from such disambiguation. … In the classic version of the ICA model (Jutten and Hérault 1991; Comon 1994; Hyvärinen et al. 2001), each observed random variable is represented as a weighted sum of independent random variables. … Cited by 4 – Related articles – All 7 versions

Domain and Discourse R Porzel – Contextual Computing, 2011 – Springer … ria (with the Baum-Welch algorithm or gradient-based methods) or maximum mutual information criteria (with … of the domain – can be employed beneficially for enhancing a dialog system’s performance on … of a hidden operator test, which was designed as a light-weight Wizard-of …

Semantic Frame-Based Spoken Language Understanding YY Wang, L Deng… – Spoken Language …, 2011 – Wiley Online Library … human-computer interaction. It has been used in many spoken language processing tasks, in particular the transactional dialogue systems, where various pieces of information need to be collected from users. A frame-based … Cited by 1 – Related articles – All 3 versions

[PDF] Innovation Engine for Blog Spaces [PDF] from A Lorincz… – 2011 – Page 1. AFRL-AFOSR-UK-TR-2011-0040 Innovation Engine for Blog Spaces Andras Lorincz Neumann János Számítógép-tudományi Társaság Eotvos Lorand University Department of Information Systems Pazmany Peter setany 1/C Budapest, Hungary H-1117 … Related articles – View as HTML – All 2 versions

Automatic Speech Recognition for ageing voices [PDF] from R Vipperla – 2011 – … Springer, 2009. (Chapter 4) • Maria Wolters, Ravichander Vipperla, and Steve Renals. Age Recognition for Spoken Dialogue Systems: Do We Need It? In Proceedings of Interspeech, Brighton, 2009. (Chapter 7) • Ravichander Vipperla, Steve Renals, and Joe Frankel. … Related articles – All 5 versions

Tandem decoding of children’s speech for keyword detection in a child-robot interaction scenario [PDF] from M Wöllmer, B Schuller, A Batliner… – ACM Transactions on …, 2011 – … inside and outside the block which control the cell through multiplicative units (depicted as small circles); input, output, and forget gate scale input, output, and internal state respectively; ai and ao denote activation functions; the recurrent connection of fixed weight 1.0 maintains … Cited by 1 – Related articles – All 3 versions

cROVER: Context-augmented Speech Recognizer based on Multi-Decoders’ Output [PDF] from MK Abida – 2011 – … Page 14. Glossary PMI Point-wise Mutual Information ROVER Recognizer Output Voting Error Reduction SCTK Speech Recognition Scoring Toolkit TN True Negative TP True Positive WA Weighted Average WER Word Error Rate WTN Minimal Cost Word Transition Network xiv … Related articles – All 5 versions

Top-Down Attention Modelling in a Cocktail Party Scenario [PDF] from ML Marchegiani – 2012 – … In Cognitive Robotics, modelling human attention can drive the development of spoken dialogue systems 1 Page 25. and help in approaching human-robot interaction issues (see, eg Lang et al. (2003) and Marchegiani et al. (2009a)). State of the Art …

Identifying prosodic prominence patterns for English text-to-speech synthesis [PDF] from L Badino – 2010 – … Page 19. Chapter 1. Introduction 7 tences or when it is integrated within a dialogue system (usually working on limited domains, eg flight booking) that provides all the necessary contextual information to the TTS system. However … Related articles – All 2 versions

Enhancement of emotion detection in spoken dialogue systems by combining several information sources [PDF] from R López-Cózar, J Silovsky… – Speech Communication, 2011 – Elsevier … experiments in the fast food domain, the negative emotional states are a consequence of failures of the dialogue system. … Note that the mutual information is positive only if P(hi | wj) > P(hi). … ie salience(wj, hi) 0) and the value 0 otherwise; cji is the connection weight between the … Cited by 3 – Related articles – All 5 versions

Sentiment analysis and opinion mining B Liu – Synthesis Lectures on Human Language …, 2012 – … Semantic Role Labeling Martha Palmer, Daniel Gildea, Nianwen Xue 2010 Spoken Dialogue Systems Kristiina Jokinen, Michael McTear 2009 Introduction to Chinese Natural Language Processing Kam-Fai Wong, Wenjie Li, Ruifeng Xu, Zheng-sheng Zhang 2009 … Cited by 2

[PDF] A PERSONALIZED RECOMMENDER AGENT FOR THE WORLD WIDE WEB-A SEMANTIC PERSPECTIVE [PDF] from PC Chang – 2010 – … User modeling (UM) plays a key role in personalization. We use the definition “A user model is a knowledge source in a natural-language dialog system which contains explicit assumptions on all aspects of the user that may be relevant to the dialog behavior of the system. … Related articles – View as HTML – All 3 versions

Bitext Alignment J Tiedemann – Synthesis Lectures on Human Language …, 2011 – … Semantic Role Labeling Martha Palmer, Daniel Gildea, and Nianwen Xue 2010 Spoken Dialogue Systems Kristiina Jokinen and Michael McTear 2009 Introduction to Chinese Natural Language Processing Kam-Fai Wong, Wenjie Li, Ruifeng Xu, and Zheng-sheng Zhang 2009 … Cited by 3 – Related articles – Library Search – All 3 versions

Computational Terminology: Exploring Bilingual and Monolingual Term Extraction [PDF] from J Foo – 2012 – … 28 4.7.2 Terms as lexicalized noun phrases . . . . . 29 4.7.3 Mutual Information and the Loglike and ?2 coefficients 29 4.7.4 C-Value/NC-value . . . . . … 34 4.8.2 Contrastive weight . . . . . 35 4.8.3 TermExtractor . . …

Learning to tell tales: automatic story generation from Corpora [PDF] from ND McIntyre – 2011 – Page 1. Learning to Tell Tales: Automatic Story Generation from Corpora Neil McIntyre T H E U NIVER S I T Y O F E DI NBU R G H Doctor of Philosophy Institute for Communicating and Collaborative Systems School of Informatics University of Edinburgh 2011 Page 2. Page 3. iii … Related articles – All 2 versions

[BOOK] Integration of World Knowledge for Natural Language Understanding E Ovchinnikova – 2012 – … Labeling task and on paraphrasing noun-noun dependencies, both of which fall out as a by-product of weighted abduction. … NLP), this type of reasoning is intended to facilitate such applications as, for example, question answering, information extraction, and dialog systems. … Related articles

Gaussian mixture optimization based on efficient cross-validation [PDF] from T Shinozaki, S Furui… – Selected Topics in Signal …, 2010 – … ap- proximated as an occupancy weighted log likelihood [4] is ex- pressed as follows: (13) (14) … 936-939. [29] LR Bahl, PF Brown, PV de Souza, and RL Mercer, “Maximum mutual information estimation of hidden Markov model parameters for speech recognition,” in Proc. … Cited by 2 – Related articles – All 6 versions

Bidirectional LSTM networks for context-sensitive keyword detection in a cognitive virtual agent framework M Wöllmer, F Eyben, A Graves, B Schuller… – Cognitive Computation, 2010 – Springer … The aim of the SEMAINE project1 is to build a Sensitive Artificial Listener-a multimodal dialogue system with the social interaction skills … input, output, and internal state respectively; ai and ao denote activation functions; the recurrent connection of fixed weight 1.0 maintains the … Cited by 18 – Related articles – All 6 versions

Combining Natural Language Processing And Statistical Text Mining: A Study Of Specialized Versus Common Languages [PDF] from J Jarman – 2011 – … 29 2.3.2 Information Theory . . . . . 29 Entropy . . . . . 31 Mutual Information . . . . . 31 InformationGain . . . . . 31 Information Gain Ratio . . . . . …

Interactive pattern recognition applied to natural language processing [PDF] from L Rodríguez Ruiz – 2010 – … Machine translation • Handwritten text recognition • Dialogue systems • etc. In spite of the fact that human language is the most natural way for us to com- municate, it is not clear how to represent and capture its relevant features to allow for an utterly automatic processing. … Cited by 1 – Related articles – All 9 versions

Emotion classification by removal of the overlap from incremental association language features JL Wu, LC Yu… – Journal of the Chinese Institute of …, 2011 – Taylor & Francis … Towards a pure Chinese spoken dialogue system. … Many researchers use pointwise mutual information (PMI) or mutual information (MI) to measure correlation confidence about the degree … anything” Another sentence, such as “This thing makes me cry and lose 10 kg weight in a … Cited by 1 – Related articles – All 4 versions

A Computational Approach to the Analysis and Generation of Emotion in Text [PDF] from F Keshtkar – 2011 – Page 1. A Computational Approach to the Analysis and Generation of Emotion in Text by Fazel Keshtkar Thesis Submitted to the Faculty of Graduate and Postdoctoral Studies in Partial Fulfillment of the Requirements For the Degree of Ph.D. of Computer Science (Ph.D.) … Related articles – All 2 versions

[PDF] Detecting Grammatical Errors with Treebank-Induced, Probabilistic Parsers [PDF] from J Wagner – 2012 – … 193 6.6 Conclusions and Future Work . . . . . 202 6.6.1 Weighted Voting with all Decision Trees . . . . . 203 6.6.2 Expand Investigation of Feature Set Combinations . . . . . 203 6.6.3 Using Probability Estimates for Accuracy Trade-Off . . . . … Cited by 1 – Related articles – View as HTML

Semi-conditional planners for efficient planning under uncertainty with macro-actions [PDF] from N Roy, R He – 2010 – Page 1. Semi-conditional Planners for Efficient Planning under Uncertainty with Macro-actions by Ruijie He BS Massachusetts Institute of Technology (2007) MS Massachusetts Institute of Technology (2008) Submitted to the … Related articles – All 2 versions

Characterizing phonetic transformations and fine-grained acoustic differences across dialects [PDF] from JP Campbell, NFY Chen – 2011 – Page 1. Characterizing Phonetic Transformations and Fine-Grained Acoustic Differences Across Dialects by MASSACHUSETTS INST OF TECHNOLOGY Nancy Fang-Yih Chen BS, National Taiwan University (2002) 8 2011 SM, National Taiwan University (2004) LIBRARIES … Related articles

Free tools and resources for Brazilian Portuguese speech recognition N Neto, C Patrick, A Klautau… – Journal of the Brazilian …, 2011 – Springer … For example, training an ASR with the maximum mutual information (MMI) cri- terion [16] requires many hours of audio data for training, otherwise the … In practice, an empirical constant is used to weight the lan- guage model probability p(T ), before combining it with the acoustic … Cited by 4 – Related articles – All 3 versions

Measuring Degrees of Semantic Opposition [PDF] from SM Mohammad, BJ Dorr, G Hirst… – 2011 – … r Understanding discourse structure and improving dialogue systems. … Our approach uses pointwise mutual information (PMI), a commonly used technique to identify semantic similarity … pair denote degrees of some variable property such as length, speed, weight, accuracy, etc … Cited by 1 – Related articles – All 2 versions

Correction of Noisy Sentences using a Monolingual Corpus [PDF] from D Chatterhee – Arxiv preprint arXiv:1105.4318, 2011 – … Such Natural Language Generation is performed by a number of applications like Machine Translation, Question Answering, Dialog Systems, 1 … They combine the edit distance approach with point- wise mutual information giving the neighbourhood co-occurrence informa- 9 … Related articles – All 3 versions

A framework for exploiting electronic documentation in support of innovation processes [TXT] from JW Uys – 2010 – Stellenbosch University Department of Industrial Engineering A Framework for Exploiting Electronic Documentation in Support of Innovation Processes JW Uys Dissertation presented for the degree of Doctor of Philosophy at Stellenbosch University. Promoter: Prof. … Cited by 3 – Related articles – All 6 versions

[PDF] Cue-Based Dialogue Act Classification [PDF] from N Webb – 2010 – … This includes the role das can play in practical applications such as dialogue systems de- velopment and evaluation. … linguistics, recent work, closely linked to the development and deployment of spoken language dialogue systems, has focused on the some of the more … Related articles – View as HTML

Mode identification using stochastic hybrid models with applications to conflict detection and resolution [PDF] from A Naseri Kouzehgarani – 2010 – Page 1. (c) 2010 by Asal Naseri Kouzehgarani. All rights reserved. Page 2. MODE IDENTIFICATION USING STOCHASTIC HYBRID MODELS WITH APPLICATIONS TO CONFLICT DETECTION AND RESOLUTION BY ASAL NASERI KOUZEHGARANI DISSERTATION … Related articles – All 4 versions

[PDF] Mixed-Initiative Clustering [PDF] from Y Huang – 2010 – … If we want to embed natural language communication in mixed-initiative clustering, it is necessary to design a dialogue system capable of discussing a clustering task, but that would divert the main focus of this study. On the other hand, a human user is more in- … Related articles – View as HTML – All 8 versions

[PDF] Towards Collaborative Process Modeling-A Framework for comparing Social Features in current BPM Tools [PDF] from J Wolf – Retrieved May, 2010 – Citeseer Page 1. Towards Collaborative Process Modeling – A Framework for comparing Social Features in current BPM Tools Master Thesis IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE (M.Sc.) IN INFORMATION SYSTEMS … Related articles – View as HTML – All 5 versions

Ontology-based Information Extraction [PDF] from C Vicient Monllaró – 2011 – Page 1. Master in Artificial Intelligence (UPC-URV-UB) Master of Science Thesis Ontology-based Information Extraction Carlos Vicient Monllaó Advisors: Antonio Moreno Ribas, David Sánchez Ruenes June, 23rd 2011 Page 2. v Agraïments … Related articles – All 3 versions

(Visited 58 times, 1 visits today)