Báo cáo khoa học: "Semisupervised condensed nearest neighbor for part-of-speech tagging" pot

... neighbor. For part-of-speech tag-ging, the error reduction over condensed nearest neighbor is more than 40%, and our model is 40%smaller than the one induced by condensed nearest neighbor. ... to condensed nearest neighbor (Hart, 1968; Alpaydin, 1997) and showedthat the algorithm leads to more condensed models,and that it performs signiﬁcantly better than con-densed nearest neighbor. ... Association for Computational Linguistics:shortpapers, pages 48–52,Portland, Oregon, June 19-24, 2011.c2011 Association for Computational LinguisticsSemisupervised condensed nearest neighbor for part-of-speech...

Báo cáo khoa học: "Semi-supervised Convex Training for Dependency Parsing" potx

... unlabeleddata resource. Our goal is to obtain better perfor-mance than a purely supervised approach withoutunreasonable computational effort. Unfortunately,although significant recent progress has ... hat-loss),which is non-convex. Therefore the objective as awhole is non-convex, making the search for globaloptimal difficult. Note that the root of the optimiza-tion difficulty for S3VMs is the non-convex ... respectively, as wementioned before. Note that in (10) we have splitthe regularizer into two parts; one for the supervisedcomponent of the objective, and the other for theunsupervised component....

Báo cáo khoa học: "Experiments on Candidate Data for Collocation Extraction" pot

... usually performed in a three-stage process(cf. Krenn (2000, 28-32) and references therein):1. The source corpus is annotated with vary-ing amounts of linguistic information (rang-ing from part-of-speech ... evaluation of methods for extract-ing collocation candidates from largetext corpora. Our research is basedon a German treebank corpus used asgold standard. Results are available for adjective+noun ... (cf.Daille (1994) and Krenn (2000)). This methodol-ogy is of little use for the evaluation of the candi-date extraction step, though, for several reasons:•The accuracy of the extraction step influencesthe...

Báo cáo khoa học: "Using Search-Logs to Improve Query Tagging" potx

... im-portant for a variety of information-retrievaltasks; however, the lack of annotated datamakes training query analysis models diffi-cult. We propose a simple, efficient proce-dure in which part-of-speech ... oftechniques for selecting snippets associated with thequery, as well as an evaluation of methods for align-ing the matching words in the query to those in theselected snippets.Specifically, for each ... car/NN”. To see why this is the case, considerthe probability for rent/VB vs rent/NNP. For rent/VBwe have 0.2 + 0.8 ×23, while for rent/NNP we have0 + 0.8 ×13assuming that π(VB|rent) =...

Báo cáo khoa học: "A HYBRID REASONING MODEL FOR INDIRECT ANSWERS" pot

... plan for answering that P~ is not going shopping tonight. Applicability conditions are necessary conditions for appropriate use of a plan operator. For example, it is inappropriate for R ... coherence rules are used as a source of knowledge for generating hypotheses. Hypothesis generation begins with ini- tializing the root of a tree of hypotheses with a proposition p0 to be related ... t is a time interval, (before tl t2) denotes that tl begins before or at the same time as t2, (app-cond q p} denotes that q is a plausible enabling con- dition for doing p, and (unless...

Báo cáo khoa học: "Rich bitext projection features for parse reranking" potx

... 0.04. For the two parses of the full sentences contain-ing the NPs in figure 1 and figure 2, we sum over7 POSParents and get a value of 0.27 for parse 1and 0.11 for parse 2. The lower value for ... probabilities considered for the two En-glish parses are p(NP|NNS, NP) for “fruit” in bothparses, p(VP|NNS, NP) for “flies” in the incorrectparse, and p(NP|NNS, NP) for “flies” in the cor-rect ... F1bycomparing hypothesized parses for the discrimi-native training set with the gold standard. To tryto find the optimal λ vector, we perform direct ac-curacy maximization, meaning that we search for the...

Báo cáo khoa học: " Contextual Analysis in Word-for-word MT" pot

... 40-41] Contextual Analysis in Word -for- word MT Kenneth E. Harper, Slavic Department, University of California, Los Angeles EXPERIMENTS with word -for- word MT of Russian scientific literature ... disadvantage of word -for- word MT, then, is its prolixity: the reader is confronted with a burdensome multi- plicity of potential equivalents (syntactic and semantic) for several words in each ... true ideoglossaries, based on actual, rather than potential, behavior. The application of contextual analysis offers great potentialities for semantic clarification. Operating again on the...

Báo cáo khoa học: "Icelandic Data Driven Part of Speech Tagging" pot

... the governingverb in a clause (non-local information). There-fore, while Icelandic case looks formally like Slavicand presents similar challenges for POS tagging, italso may be syntactically-determined, ... this new information.If an incorrect hypothesis is selected during train-ing, the algorithm promotes the score of the correcthypothesis and demotes the selected one. See Shenet al. for a detailed ... Joshi (1999).3 Icelandic MorphologyIcelandic is notable for its morphological richness.Verbs potentially show as many as 54 differentforms depending on tense, mood, voice, person and33number....

Báo cáo khoa học: "An Automatic Generation System for Grammar Tests" pot

... intend _______ you that we cannot approve your application. (A) to inform (B) to informing (C) informing (D) informFigure 4: An example of generated question.Figure 5: An example of ... possessive case of the personal noun in English. The possessive form comprises an apostrophe and the letter “s”. For example, the possessive form of the personal noun “Mary” is “Mary’s”. The grammatical ... number of patterns for the development of structural tests. For example, a verb-related pattern requiring an infinitive as the complement (e.g., the verb “tend”) can be formed from the sentence...

Báo cáo khoa học: "A Generative Lexicon Perspective for Adjectival Modification" potx

... into on- tological domains for modelling some forms of metaphors. • Types are not sufficiently 'constrained' to ac- count for the constraints holding, for each predicate, on the ... in particular for the Telic role, (explored e.g. in the EuroWordNet project, the European WordNet). Qualias are well-designed and useful for nouns, but look more artificial for other lexical ... of metonymies: 'image for person, image be- ing a part of a person' (a good reputation), 'tool for profession' (a good scalpel), 'place for profession' (...

Xem thêm