Báo cáo khoa học: "An Unsupervised Dynamic Bayesian Network Approach to Measuring Speech Style Accommodation" doc

... describethe speech features used, and the proposed mod-els.3.1 Dynamic Bayesian NetworksThe theory of Bayesian networks is well doc- umented and understood (Jensen, 1996; Pearl,1988). A Bayesian network ... 27 2012.c2012 Association for Computational LinguisticsAn Unsupervised Dynamic Bayesian Network Approach to Measuring Speech Style AccommodationMahaveer Jain1, John McDonough1, Gahgene ... attempt to capture the dependence of speech features on speaking style, as well as theeffect of persistence and accommodation on style. We use a dynamic Bayesian network (DBN) for-malism to capture...

Báo cáo khoa học: "Distributed Listening: A Parallel Processing Approach to Automatic Speech Recognition" pot

... of Distributed Listening research is to take a unique approach in order to enhance the success of the traditional approaches to speech recognition. The approach of Distributed Listen-ing directly ... with multiple speech recognizers in an effort to improve speech recognition, as discussed next. 2.1 Enhanced Majority Rules Barry (et al., 1994) took three different Automatic Speech Recognition ... resolution algorithm to resolve conflicts. Figure 1. Distributed Listening Architecture 1732 Background Automatic speech recognition systems convert a speech signal into a sequence of words,...

Tài liệu Báo cáo khoa học: "An Unsupervised Model for Joint Phrase Alignment and Extraction" ppt

... HIER ﬁrst attempts to generate thefull sentence as a single phrase from Pt, then fallsback to ITG -style derivations to cope with sparsity.We allow for this within the Bayesian ITG contextby ... distribution with the averagelength parameter λ. As long phrases lead to spar-sity, we set λ to a relatively small value to allowus to bias against overly long phrases4. Pm1is theword-based Model ... according to the model. By choosingthis probability appropriately, we can incorporateprior knowledge of what phrases tend to be aligned to each other. We calculate Pbaseby ﬁrst choosingwhether to...

Tài liệu Báo cáo khoa học: "An Unsupervised Approach to Recognizing Discourse Relations" pdf

... word pairs that are not likely to be good predictorsof discourse relations. To test this hypothesis, we decided to carry outa second experiment that used as predictors onlya subset of the word ... us access to general purpose knowledgebases that assert that “similar(libya,rwanda)”.The approach we advocate in this paper is in somerespects less ambitious than current approaches to discourse ... NO-RELATION-SAME-TEXT. Andwe extracted automatically 1,000,000 examples ofwhat we hypothesize to be cross-document non-relations, by randomly selecting two sentences fromdistinct documents. As in the case...

Báo cáo khoa học: "An Unsupervised Approach to Prepositional Phrase Attachment using Contextually Similar Words" potx

... unsupervised corpus-based approach to prepositional phraseattachment that outperforms previous unsupervised techniques and approaches theperformance of supervised methods. Unlikeprevious unsupervised ... University of Manitoba.AbstractPrepositional phrase attachment is acommon source of ambiguity in naturallanguage processing. We present an unsupervised corpus-based approach to prepositional ... performance to supervisedmethods. Unlike previous unsupervised approaches in which training data isobtained by heuristic extraction ofunambiguous examples from a corpus, weuse an iterative process to...

Báo cáo khoa học: "An Unsupervised Morpheme-Based HMM for Hebrew Morphological Disambiguation" pdf

... is similar to words extraction in automatic speech recogni-tion (ASR)(Jurafsky and Martin, 2000, chp. 5,7).The states of the ASR model are phones, whereeach observation is a vector of spectral ... supervised training of a speech recognizer combines a training corpus of speech wave ﬁles, together with word-transcription, andlanguage model probabilities, in order to learn thephones model.There ... algorithm to our text rep-resentation for ﬁrst-order HMM, which works inO(˙T ) time.4 Experimental ResultsWe ran a series of experiments on a Hebrew corpus to compare various approaches to the...

Báo cáo khoa học: "an Unsupervised Web Relation Extraction System" pot

... target concepts in a set of documents. The sys-tem then learns a model of the extraction task, which can be applied to new documents auto-matically. Both of these approaches require massive ... or more arbitrary tokens, and slots match instance attributes. The limited skips match zero or more arbitrary tokens, which must not belong to entities of the types equal to the types of the ... that do not belong to any of the entity types above. In order to do the extraction, the patterns gener-ated by the Pattern Learner are converted to the TEG syntax and added to the pre-built...

Báo cáo khoa học: "An Unsupervised System for Identifying English Inclusions in German Text" doc

... data.Domain Tokens Types TTRInternet Total 15919 4152 0.26English 963 283 0.29Space Total 16066 3938 0.25English 485 73 0.15EU Total 16028 4048 0.25English 49 30 0.61Table 1: English token and ... required to recognise them. Our aimis to address these issues in future work.4 System DescriptionOur system is a UNIX pipeline which convertsHTML documents to XML and applies a set of mod-ules to ... sparse in the EU domain (49 tokens) butconsiderably frequent in the documents in the inter-net and space travel domains (963 and 485 tokens,respectively). The type-token ratio (TTR) signalsthat...

Báo cáo khoa học: "An Unsupervised Model for Statistically Determining Coordinate Phrase Attachment" pptx

... salad of lettuce and tomatoes 'Lettuce' attaches low to 'tomatoes', giving: L-attach: (salad (of ((lettuce) and (tomatoes))) [AR98] models. In addition to these, a corpus- ... helpful to compare its performance to that of an analogous supervised model. The smaller the error reduction when going from unsupervised to supervised models, the more comparable the unsupervised ... from the unsupervised model presented here to the backed-off model is 13%. This is comparable to the 14.3% error reduction found when going from JAR98] to [CB95]. It is interesting to note...

Báo cáo khoa học: "An Unsupervised Vector Approach to Biomedical Term Disambiguation: Integrating UMLS and Medline" ppt

... con-cept vector is the closest to the instance vector to thetarget word. To create the the instance vector, we use the wordsthat occur in the same abstract as the target word asfeatures. To create ... paper, we introduce an unsupervised vector approach to disambiguate words in biomedical textusing contextual information from the UMLS andMedline. We compare our approach to Humphrey etal. (2006) ... USAbthomson@cs.umn.eduAbstractThis paper introduces an unsupervised vector approach to disambiguate words in biomedi-cal text that can be applied to all-word dis-ambiguation. We explore using contextualinformation...

Xem thêm