Font Size: a A A

Desigh And Implement Of Parser Based On Grammar Function And Collocation

Posted on:2016-08-18Degree:MasterType:Thesis
Country:ChinaCandidate:L F WuFull Text:PDF
GTID:2348330512972855Subject:Information Science
Abstract/Summary:PDF Full Text Request
Parsing is an important part in the field of Information Process,which also is a foundational technology.The result of parsing is directly associated with a series of works,such as machine translation,information extraction,text abstract,information filtration and so on.The ultimate goal of this article is to design and implement an automatic parser,which is based on the principle of grammar function and collocation.Nowadays,most parsers are designed to parsing according to the part of speech tagging of word sequence,these analytical methods not only increase ambiguity,and the final result can't provide abundant and accurate syntactic information.In view of this,this paper applies the idea of grammar function matching to accomplish the procedure of natural language's parsing.There are two important part of automatic parsing,the one is knowledge base of grammar function,the other one is the algorithm of parsing.This paper's knowledge base of grammar function is built on the adjusted TCT 973,which included words' grammar functions and grammar rules,these grammar rules could guide the transition between grammar functions.The algorithm of parsing referred to chart parsing,a widely applied parsingalgorithm,and parsing a sentencebottom-up.This paper divided the task of parsing into two subtasks:tagging words' grammar functions automatically and then parsing automatically.The section of grammar functions tagging imitated the processing mode of part of speech tagging,and tagged with Hidden Markov Model(HMM),which eventually obtain good results.Besides this,we import the corpus of collocation creativity.With the guide of collocation corpus,we rearrange a sentence which cann't be parsed successfully.The procedure of rearrangement is combine two words which are common collocation in our collocation corpus into one new word,this processing can reduce the dependency of uncommon rules,meanwhile accomplish error recovery.At last,we compare ours method with other existed method,the result shows that precision and F1-measure both have a certain degree of improvement.
Keywords/Search Tags:parsing, grammar function matching, collocation, HMM
PDF Full Text Request
Related items