Tài liệu Báo cáo khoa học: "Bayesian Word Sense Induction" pdf

... Introduction Sense induction is the task of discovering automat-ically all possible senses of an ambiguous word. Itis related to, but distinct from, word sense disam-biguation (WSD) where the senses ... the generated word. Then, the word sense is selected based on the word, neighbor, and topic.Boyd-Graber et al. (2007) extend the topic mod-eling framework to include WordNet senses as alatent ... sam-pling a sense from the sense distribution, thenchoosing a word from the sense- context distribu-tion. P(si= j) denotes the probability that the jth sense was sampled for the ith word token...

Tài liệu Báo cáo khoa học: "Direct Word Sense Matching for Lexical Substitution" ppt

... all WordNet synonyms of the target word, underall its possible senses, and picking randomly oneof the synonyms as the source word. For example,the word ‘disc’ is one of the words in the Sense- val ... were excluded sincetheir sense annotation in Senseval-3 is not basedon WordNet senses. The Senseval dataset includesa set of example occurrences in context for each word, split to training and ... of the 25 words inthe Senseval sample as a target word for the sense matching task. Next, we had to pick for each target word a corresponding synonym to play the role ofthe source word. This...

Tài liệu Báo cáo khoa học: "SenseLearner: Word Sense Disambiguation for All Words in Unrestricted Text" doc

... describes SENSELEARNER – aminimally supervised word sense disam-biguation system that attempts to disam-biguate all content words in a text usingWordNet senses. We evaluate the accu-racy of SENSELEARNER ... modellearning Sense taggedtextsemantic modelsSenseLearnerdefinitions Word sense disambiguationTrained semanticmodels Sense taggedtextsFigure 1: Semantic model learning in SENSE- LEARNERtagged ... polyse-mous word within a given context. Applications suchas machine translation, knowledge acquisition, com-mon sense reasoning, and others, require knowledgeabout word meanings, and word sense...

Tài liệu Báo cáo khoa học: "Bayesian Symbol-Reﬁned Tree Substitution Grammars for Syntactic Parsing" pptx

... grammars with only unary and binaryproductions. We replace lexical words with count≤ 5 in the training data with one of 50 unknownwords using lexical features, following (Petrov et al.,2006). ... of(a). The reﬁnement annotation is hyphenated with a nonterminal symbol.morphology analysis, word segmentation (Johnsonand Goldwater, 2009), and dependency grammar in-duction (Cohen et al.,...

Tài liệu Báo cáo khoa học: "Improving Word Representations via Global Context and Multiple Word Prototypes" pdf

... multi-prototype VSM where word sense discrimination is ﬁrst applied by clus-tering contexts, and then prototypes are built usingthe contexts of the sense- labeled words. However, inorder to ... for clustering word instances, which is used in the multi-prototype ver-sion of our model that accounts for words with mul-tiple senses.We evaluate our new model on the standardWordSim-353 (Finkelstein ... context and one represen-tation per word. This is problematic becausewords are often polysemous and global con-text can also provide useful information forlearning word meanings. We present a newneural...

Tài liệu Báo cáo khoa học: "Unsupervized Word Segmentation: the case for Mandarin Chinese" doc

... densitydistributions for words vs. non-words, we observedthat the VBE at both boundaries were the most dis-criminative value. Therefore, we decided to take inaccount the VBE only at the word- candidate ... corresponding to thesequence of words w0w1. . . wm, and len(wi) is thelength of a word wiused here to be able to com-pare segmentations resulting in a diﬀerent numberof words. This best segmentation ... hypothesis and consider com-plex morphological word structures, we expect alarge VBE at the boundaries of interesting units andmore unstable variations inside “words.” This expec-tation was conﬁrmed...

Tài liệu Báo cáo khoa học: "Enhanced word decomposition by calibrating the decision threshold of probabilistic models and using a model ensemble" pdf

... relative word positions andfound out that the calibrated PROMODES-H pre-dicted non-boundaries better for initial word posi-tions whereas the calibrated PROMODES for mid-and final word positions. ... term word morphology. It is worthwhilestudying this internal structure since a languagedescription using its morphological formation ismore compact and complete than listing all pos-sible words. ... morphological analy-sis: word decomposition into morphemes, build-ing morpheme dictionaries, defining morphosyn-tactical rules which state how morphemes canbe combined to valid words and defining mor-phophonological...

Tài liệu Báo cáo khoa học: "Learning Word-Class Lattices for Deﬁnition and Hypernym Extraction" doc

... frequent words F to generalizewords to word classes”. We deﬁne a word classas either a word itself or its part of speech. Givena sentence s = w1, w2, . . . , w|s|, where wiis thei-th word ... order of symbols like in word/ phoneme lat-tices, and nodes are clusters of salient words ag-gregated using synonymy, similarity, or subtreesof a thesaurus. However, salient word selectionand aggregation ... where wiis thei-th word of s, we generalize its words wito word classes ωias follows:ωi=wiif wi∈ FP OS(wi) otherwisethat is, a word wiis left unchanged if it occursfrequently...

Tài liệu Báo cáo khoa học: "Learning Word Vectors for Sentiment Analysis" ppt

... com-pare our model’s word representations with severalbag of words weighting methods, and alternative ap-proaches to word vector induction.4.1 Word Representation LearningWe induce word representations ... simi-larity of w with all other words w′, we can ﬁnd thewords deemed most similar by the model.Table 1 shows the most similar words to givenquery words using our model’s word representationsas ... continu-ous similarities between words as distance or anglebetween word vectors in a high-dimensional space.The general approach has proven useful in taskssuch as word sense disambiguation, named...

Tài liệu Báo cáo khoa học: "Joint Word Segmentation and POS Tagging using a Single Perceptron" docx

... last word can be a complete word ora partial word. A problem arises in whether to givePOS tags to incomplete words. If partial words aregiven POS tags, it is likely that some partial wordsare ... words12the ending characters c1and c2of two con-secutive words13a word of length l with previous word w14a word of length l with next word wTable 1: Feature templates for the baseline segmentor2 ... Chinese, are shown in Table 2.The word segmentation features are extractedfrom word bigrams, capturing word, word lengthand character information in the context. The word length features are normalized,...

Xem thêm