Báo cáo khoa học: "Unsupervised Part-of-Speech Tagging with Bilingual Graph-Based Projections" doc

Báo cáo khoa học: "Dialogue Act Tagging with Transformation-Based Learning" docx

... Word n-grams Entropy minimization Entropy minimization with clustering Entropy minimization with filtering Entropy minimization with filtering and clustering # 0 936 16271 1053 1029 ... tag. Brill (1995b) developed an unsupervised version of TBL for Part-of-Speech Tagging, but this algorithm must be initialized with words that can be tagged unambiguously, 13 and in discourse, ... of dialogues with dialogue 1150 acts, the rules are applied, in turn, to every utterance in the corpus, and each utterance that satisfies the conditions of a rule is relabeled with that...

Báo cáo khoa học: "Bitext Dependency Parsing with Bilingual Subtree Constraints" docx

... Perl.23ateHeHe:1:2-ate:2:0atemeatate:1:0-meat:2:1ate with ate:1:0 -with: 2:1meatthethe:1:2-meat:2:0 with fork with: 1:0-fork:2:1forkaa:1:2-fork:2:0Figure 4: Examples of bigram-subtreesatemeat with ate:1:0-meat:2:1 -with: 3:1ate with ... theparser with the monolingual features monolingualparser.3.2 Parsing with bilingual featuresIn this paper, we parse source sentences with thehelp of their translations. A set of bilingual ... ate:1:0-NULL:2:1-meat:3:1the:1:3-NULL:2:3-meat:3:0a:1:3-NULL:2:3-fork:3:0 with: 1:0-NULL:2:1-fork:3:1ate:1:0-the:2:3-meat:3:1 ate:1:0 -with: 2:1-fork:3:2 with: 1:0-a:2:3-fork:3:1NULL:1:2-He:2:3-ate:3:0He:1:3-NULL:2:1-ate:3:0 ate:1:0-meat:2:1-NULL:3:2ate:1:0-NULL:2:3 -with: 3:1 with: 1:0-fork:2:1-NULL:3:2NULL:1:2-a:2:3-fork:3:0a:1:3-NULL:2:1-fork:3:0ate:1:0-NULL:2:3...

Báo cáo khoa học: "Unsupervised Word Alignment with Arbitrary Features" potx

... jump conjoined with an indicator feature for the target length n, the dis-cretized jump feature conjoined with the class of saj,and the discretized jump feature conjoined with theclass ... California.K. Gimpel and N. A. Smith. 2009. Cube summing, ap-proximate inference with non-local features, and dy-namic programming without semirings. In Proc. ofEACL.N. Habash and F. Sadat. 2006. Arabic ... conclude with an analysis of relatedprior work (§6) and a general discussion (§8).2 ModelIn this section, we develop a conditional modelp(t | s) that, given a source language sentence s with length...

Báo cáo khoa học: "Unsupervised Relation Discovery with Sense Disambiguation" docx

... the “music” sense.Document theme: Sometimes, the same patterncan express different relations in different docu-ments, depending on the document’s theme. Forinstance, in a document about politics, ... argument areadded with prefix “lc:”. Similarly the two words tothe right of the destination arguments are added with prefix “rc:”.Each document in the NYT corpus is associated with many descriptors, ... LDA model to find the theme of adocument. See Section 3.1 for details.Sentence theme: A document may cover severalthemes. Moreover, sometimes the theme of a doc- ument is too general to disambiguate...

Báo cáo khoa học: "Unsupervised Sense Disambiguation Using Bilingual Probabilistic Models" pdf

... an abundance of bilingual documents andmany more can potentially be mined from the web.It should be possible using our approach to (noisily)assign sense tags to words in such documents, thusproviding ... approaches to word-sense disambiguation. For example, the use of par-allel corpora for sense tagging can help with wordsense disambiguation (Brown et al., 1991; Dagan,1991; Dagan and Itai, 1994; ... to sentence (with a shared mean-ing), while the data we are modeling are alignedsingle-word translations , in which the in-tended meaning of does not always match per-fectly with the intended...

Báo cáo khoa học: "Flexible Answer Typing with Discriminative Preference Ranking" docx

... binary classification simply deals with theis/is-not relationship and will have difficulty whentwo responses with similar feature values are clas-sified differently. With preference ranking, viola-tions ... appropriatecandidates.4 ExperimentsTo compare with the prior approach of Pinchakand Lin (2006), we use a set of what and whichquestions with question focus (questions with anoun phrase following the wh-word). ... appropriate eventhough only Calgary is correct. This task comes with a moderate degree of difficulty, especiallywhen dealing with questions for which appropriateanswers are less obvious (such...

Báo cáo khoa học: "Discriminative Sentence Compression with Soft Syntactic Evidence" doc

... sentences.The primary problem with this model is that mostof the model features encode properties related toincluding or dropping constituents from the tree with no encoding of bigram or trigram ... analysis ofthe sentence. To do this we parse every sentencetwice, once with a dependency parser (McDon-ald et al., 2005b) and once with a phrase-structureparser (Charniak, 2000). These parsers have ... calculat-ing model parameters.1 IntroductionThe ability to compress sentences grammatically with minimal information loss is an importantproblem in text summarization. Most summariza-tion...

Báo cáo khoa học: "Generating referring expressions with a unification grammar" doc

... the value of feature F; as usual in Prolog, symbols starting with a lower-case letter are con- stants, while symbols starting with an upper-case letter are variables. Focus and mention vectors ... problem of updating mention ratios by unification, without resorting to statements like Df is Di + 1. This can be done by replacing numbers with lists of the appropriate length, so that for ... 0/2 is represented by the term [] / [_, _] With this convention, the relationship between the mention ratios of F(UN) and F(uo) can be stated without an accompanying numerical con- straint:...

Báo cáo khoa học: "Unsupervised Part-of-Speech Tagging with Bilingual Graph-Based Projections" doc

... 19-24, 2011.c2011 Association for Computational LinguisticsUnsupervised Part-of-Speech Tagging with Bilingual Graph-Based ProjectionsDipanjan Das∗Carnegie Mellon UniversityPittsburgh, ... Ryan McDonald. 2011.A universal part-of-speech tagset. ArXiv:1104.2086.Sujith Ravi and Kevin Knight. 2009. Minimized modelsfor unsupervised part-of-speech tagging. In Proc. ofACL-IJCNLP.Libin ... oracles in addition to two variantsof our graph-based approach. We were intentionallylenient with our baselines:• EM-HMM: A traditional HMM baseline, with multinomial emission and transition...

Tài liệu Báo cáo khoa học: "Unsupervised Semantic Role Induction with Global Role Ordering" doc

... expected semantic role se-quence. Sentence a is in active voice with sequence(A0, P REDICA T E, A1) and sentence b is in passivevoice with sequence (A1, P R EDICAT E, A0). Addi-tional global preferences, ... argument instances.We chose a uniform Dirichlet prior with concentra-tion parameter as 0.1 for all the model parametersin Algorithm 1 (set roughly, without optimization1).50 training iterations ... 83.68 78.23 80.874 83.72 78.08 80.80Table 2: Performance variation with the number ofPRs (excluding ST ART , END and P RED) With only this additional ordering information, theperformance is...

Xem thêm