Báo cáo khoa học: "Minimized models and grammar-informed initialization for supertagging with highly ambiguous lexicons" pdf

... creating grammars and resources for CCGparsers for new domains and languages is highly labor- and knowledge-intensive. Baldridge (2008)uses grammar-informed initialization for HMMtag transitions ... Association for Computational Linguistics, pages 495–503,Uppsala, Sweden, 11-16 July 2010.c2010 Association for Computational LinguisticsMinimized models and grammar-informed initialization for supertagging ... starting with uniform transition distributions.We refer to a model learned from basic EM (uni-formly initialized) as EM, and to a model with grammar-informed initialization as EMGI.4 Minimized models...

Báo cáo khoa học: "Minimized Models for Unsupervised Part-of-Speech Tagging" pot

... methods and results has accumulatedaround it, and these can be clearly compared with one another.We use the standard test set for this task, a24,115-word subset of the Penn Treebank, for which ... Southern CaliforniaInformation Sciences InstituteMarina del Rey, California 90292{sravi,knight}@isi.eduAbstractWe describe a novel method for the taskof unsupervised POS tagging with a dic-tionary, ... the ACL and the 4th IJCNLP of the AFNLP, pages 504–512,Suntec, Singapore, 2-7 August 2009.c2009 ACL and AFNLPMinimized Models for Unsupervised Part-of-Speech TaggingSujith Ravi and Kevin...

Tài liệu Báo cáo khoa học: "Semantic Information and Derivation Rules for Robust Dialogue Act Detection in a Spoken Dialogue System" pptx

... PST as a knowl-edge source for DA detection, we essentially need tomodel the relationship between the random DA and the random DR. Denote the random DA by X and the random DR by Y . Given a text ... z-score (Larsen and Marx,2000) of each word w in the ASR output. Figure 3illustrates the PST for the sentences: Where is theAnping-Fort. There are two keywords Where and Anping-Fort and two non-keywords ... Association for Computational Linguistics:shortpapers, pages 603–608,Portland, Oregon, June 19-24, 2011.c2011 Association for Computational LinguisticsSemantic Information and Derivation Rules for...

Tài liệu Báo cáo khoa học: "Fast Decoding and Optimal Decoding for Machine Translation" doc

... position q with probabilityd(q–k class(f )), and so forth.NULL-generated. After heads and non-headsare placed, NULL-generated words are permutedinto the remaining vacant slots randomly. If ... is chosen with probability 1/ .These stochastic decisions, starting with e, re-sult in different choices of f and an alignment of f with e. We map an e onto a particulara,f pair with probability:P(a, ... solve and http://www.cplex.com.Figure 3: A salesman graph for the input sen-tence f = “CE NE EST PAS CLAIR .” There isone city for each word in f. City boundaries aremarked with bold lines, and...

Tài liệu Báo cáo khoa học: "Combining Functionality and Object Orientedness for Natural Language Processing" ppt

... composite expressions are constructed only with a binary form of function application. Thus, if z and I/ are well-formed formulas of PAL, so is a form z(y). Expressions of PAL are related to ... descendants, say z and y, a PAL expression for the phrase is one of the following forms: <z>( <V>) or <p>( <z>) where ~a> stands for a PAL expression for a phrase ~*. ... Combining Functionality and Ob]ec~Orientedness for Natural Language Processing Toyoakl Nishida I and Shuji Doshita Department of Information Science, Kyoto University Sakyo-ku,...

Tài liệu Báo cáo khoa học: "Learning Parse and Translation Decisions From Examples With Rich Context" pdf

... surface and lexi- cal form, syntactic and semantic category, subframes with syntactic and semantic roles, and form restric- 1The lexicon provides part-of-speech information and links words to ... describes the syntactic and semantic role structures for verbs, with currently 242 entries. The following representative examples, for easier understanding rendered in English and not in fea- ture ... that let our system start with the correct parse, effectively just testing the transfer and generation module. The resulting translations, in randomized order and without iden- tification,...

Báo cáo khoa học: "Joint Processing and Discriminative Training for Letter-to-Phoneme Conversion" ppt

... approaches, such as those of Bisani and Ney (2002) and Marchand and Damper (2000), haveshown that better performance can be achieved bypairing letter substrings with phoneme substrings,allowing ... classiﬁed intosubstrings of phonemes, and introducing an inputsegmentation step before prediction begins. Themethod accounts for sequence information with post-processing: the numerical scores ... trees (Daelemans and Bosch,1997; Black et al., 1998) or instance-based learn-ing (Bosch and Daelemans, 1998). These systemspredict a phoneme for each input letter, using theletter and its context...

Báo cáo khoa học: "Syntactic Features and Word Similarity for Supervised Metonymy Resolution" pot

... (Lakoff and Johnson, 1980; Fass, 1997;Stern, 1931)). In a place -for- people pattern,a place stands for any persons/organisations associ-ated with it, e.g., for sports teams in (2), (3), and (4), and ... evi-dence for any reading and lowering both recall and precision. Bad predictors are the verbs “tohave” and “to be” and some prepositions suchas with , which can be used with metonymic(talk with ... and Agirre, 2000): 0.1 is added to both the de-nominator and numerator.The target readings to be distinguished areliteral, place -for- people, place -for- event, place -for- product, othermet and mixed....

Báo cáo khoa học: "Plot Induction and Evolutionary Search for Story Generation" pot

... Traditionally, most of this knowl-edge is created by hand, and the effort must be re-peated for new domains, new characters and plotelements.Fortunately, recent work in natural languageprocessing ... coherence) that is good enough for peo-ple to understand. Cheng and Mellish (2000) focuson the interaction of aggregation and text planning and use genetic algorithms to search for the bestaggregated ... players’actions (Barros and Musse, 2007); and assistingteachers to create or personalize stories for theirstudents (Riedl and Young, 2004).A major stumbling block for the widespread useof...

Báo cáo khoa học: "Phonological Constraints and Morphological Preprocessing for Grapheme-to-Phoneme Conversion" pot

... for Pronunciation byAnalogy (PbA) (Marchand and Damper, 2005) onthe same corpus. Marchand and Damper reported aprocessing time of several hours for c4), two days for c2) and several days for ... that uses the states without stress and syllable flags and enforces the constraints byeliminating invalid transitions, cf. section 3.1) tooutperform the integrated version, v1, and more sig-nificantly ... component for syllabification and one for word stress assignment.The model for syllabification computes the mostprobable sequenceˆbn1of syllable boundary-tagsˆb1,ˆb2, ,ˆbn for a given...

Xem thêm