Báo cáo khoa học: "Generalizing Tree Transformations for Inductive Dependency Parsing" potx

... MS.2.3 Transformations, Parsers, and TreebanksPseudo-projective parsing and transformations for coordination and verb groups are instances of thesame general methodology:1. Apply a tree transformation ... 968–975,Prague, Czech Republic, June 2007.c2007 Association for Computational LinguisticsGeneralizing Tree Transformations for Inductive Dependency ParsingJens Nilsson∗Joakim Nivre∗†∗V¨axj¨o ... compari-son to the other treebanks, making the inverse trans-formation for PADT more problematic than for theother treebanks. The absence of a positive effect for PADT is therefore understandable...

Báo cáo khoa học: "Discriminative Classiﬁers for Deterministic Dependency Parsing" docx

... conclusions in section 6.2 Inductive Dependency ParsingThe system we use for the experiments uses nogrammar but relies completely on inductive learn-ing from treebank data. The methodology ... ☎❄NMODNNmarkets✞ ☎❄PMOD..❄✞☎PFigure 1: Dependency graph for an English sentence from the WSJ section of the Penn Treebankrequire non-projective dependency arcs. For thepurpose of this paper, however, ... thatSVM outperforms MBL for the three most com-plex models Φ3, Φ4and Φ5, both for English andChinese. The results for Swedish are less clear,although the labeled accuracy for Φ3and Φ5aresigniﬁcantly...

Tài liệu Báo cáo khoa học: "Web-Scale Features for Full-Scale Parsing" doc

... ofPenn Treebank into training (sections 2-21), devel-opment (section 22) and test (section 23). We usedthe ‘pennconverter’5tool to convert Penn trees fromconstituent format to dependency format. ... errorreduction. Results are for dependency parsing on the dev set for iters:5,training-k:1.tal errors break down by gold head. For example,the 12.1% total error reduction for attachments of anIN ... compute features for the full parsingtask. To show end-to-end effectiveness, we incor-porate our features into state-of-the-art dependency and constituent parsers. For the dependency case,we...

Tài liệu Báo cáo khoa học: "A rich environment for experimentation unification grammars" potx

... into discussion of the under- lying complexity, for which we unfortunately do not have space here, we give an exter- nal form of a lexical entry for some of the senses of the German verb traeumen. ... PATR-style languages is their lack of facilities for defining new abstractions and expressing linguistic generalisations not foreseen (or even foreseeable) by the language designer. This becomes ... subscripts. Three other special symbols are used: + stands for the unifica- tion operator * stands for top, the underdefined element. # stands for bottom, the overdefined element that corresponds...

Tài liệu Báo cáo khoa học: "A COMMON FRAMEWORK FOR ANALYSIS AND GENERATION" potx

... compo- nent describing their functional structure. For the rule for S, for instance, we add an ele- ment describing what the SUBJECT and PRED are for. We could say that the SUBJECT is the theme ... complete formal paraphrase of (1) must include at least as much information as we have given above. In particular, the logical structure of our para- phrases contains essential information ... consid- erable debate about the best way of representing this information. 2 GENERATION Suppose we have the formula given above as a for- mal paraphrase of (1), and we want to generate an English...

Báo cáo khoa học: "Using Derivation Trees for Treebank Error Detection" potx

... SheikhNP#a3qmpsummit#b1S:1.2NPNP^$rmSharm#b2NPAl$yxThe Sheikh For (2a) For (2b) For (2c) A:1.1,left#b3NPAl$yxThe SheikhA:1.1,left#c3$rmSharmFigure 1: Etrees and Derivation Trees for (2abc). tree- substitution, sister-adjunction, ... “building blocks” of the full trees of a treebank.This decomposition of the full tree into etrees alsoresults in a “derivation tree that records how the el-ementary trees relate to each other.We ... the etreesand derivation trees as just described require reﬁne-ment to be truly appropriate for comparing nuclei.The reason is that etrees might encode more infor-mation than is relevant for...

Báo cáo khoa học: "A Bayesian Model for Unsupervised Semantic Parsing" ppt

... arguments are lo-cal in the dependency tree; that is, one lexical itemcan be a semantic argument of another one only ifthey are connected by an arc in the dependency tree. This is a slight simpliﬁcation ... syntactic realizations for each semantic class and distributions of syntactic dependency arcs for argument types. In this sectionwe present relevant background on PY processes. For a more detailed ... word w uniformly from thetreebank distribution, then we decide on the num-ber of dependents of w using the geometric distribu-tion Geom(q(C)). For every dependent we generatea dependency...

Báo cáo khoa học: "Choosing Sense Distinctions for WSD: Psycholinguistic Evidence" potx

... Response times for distantly related pairs was faster than for homo-nyms (distantly related mean: 1253ms, homonym mean: 1406ms; t32=2.38; p<.0001). Accuracy was en-hanced as well for this ... implications for automatic word sense disam-biguation. 1 Introduction* The problem of creating a successful word sense disambiguation system begins, or should begin, well before methods ... accessibility, has become the resource of choice for WSD. However, some have questioned whether WordNet’s fine-grained sense distinctions are ap-propriate for the task (Ide & Wilks, 2007; Palmer...

Báo cáo khoa học: "A Connectionist Architecture for Learning to Parse" potx

... computed. However the link weights that perform this computation are the same for all the positions in the sequence. Therefore the infor- mation that was learned for an input and context in one sequence ... pattern for the gestalt inputs is straightforward, since these inputs per- tain to information about the sentence as a whole. Whenever a tag is input to the network, the activa- tion pattern for ... averaged the performance of the best two networks to obtain the figures in the second row of table 1. For experiment (3), labeled parse trees were used as a target output, i.e. for each word...

Báo cáo khoa học: "Learning PP attachment for filtering prosodic phrasing" potx

... according to these search meth-ods was used for freq(X P). The freq(X) was de-rived from the WWW by performing a separatesearch for the single word form.Method I: cooccurrence NnearP or VnearPVolk ... al-gorithm for Dutch (van Herwijnen and Terken,2001b). This phrasing algorithm uses syntactic in-formation and sentence length for the allocationof prosodic phrase boundaries. For a subset ... syntac-tic treebank. Collection of this data is described inSection 2. We extract lexical head features (uni-gram and bigram) from the treebank occurrences,and enrich this data with cooccurrence informa-tion...

Xem thêm