Font Size: a A A

Studies On Prepositional Phrase Boundary Identification Based On Usage Attribute

Posted on:2012-06-18Degree:MasterType:Thesis
Country:ChinaCandidate:Y C YuanFull Text:PDF
GTID:2218330338458185Subject:Computer software and theory
Abstract/Summary:PDF Full Text Request
Because positions are belong to function words, they do not have specific meanings like notionals word. However, prepositional phrases which are constructed prepositional and guided objects have important meanings in syntax analysis. Prepositional phrases structure is the most common phrase structure in Chinese text. So prepositional phrase boundary identification is one of the important contents in natural language processing. This paper attempts to identificate prepositional phrase boundary based on rules and statistics on the basis of preposition knowledge base. From on the experimental results, the result of statistical method is higher 23% than rule method on average. Identification results are different for different statistical models,such as ME model is best, the second is CRF model, SVM model is worse. In addition,the overall recognition accuracy of th ree statistical models have improved after joining in preposition usage attribute features, ME increased 0.14%, CRF increased 0.01%, SVM increased 0.22%. Experimental results show that prepositional usage attribute overall is help foe prepositional phrases boundary recognition. It also shows that prepositional phrase boundary identification based on prepositional usage attribute is effective.The key work performed in this paper include:(1) According to the thoughts of builting the "Trinity" knowledge base of mordern Chinese functional words which are put forward by YuShiWen and preposition knowledge base framework which is established by PengShuang and ZanHongYing, this peper perfected the prepositional knowledge base.(2) This paper convered preposition rules in prepositions knowledge base to static rules of prepositional phrase boundary identification. Then it identificated prepositional phrases boundary using these static rules.(3) Proofreaded prepositional usages of prepositional corpora, marked all correct prepositional phrase structureto get experimental corpora which contain prepositional usage attribute and accurate prepositional phrase boundary markers. Then we inspected prepositional phrase boundary identification results on the influence of prepositional usage attribute, validated whether prepositional usage attribute is effective for prepositional phrase boundary identification.Finally, this paper summarized the research work. It pointed out that prepositional knowledge base is helpful for prepositional phrase boundary identification. Then it determined future research according to this study results.
Keywords/Search Tags:Prepositional Usage, Prepositional Phrase Boundary Identification, Maximum Entropy, Conditional Rondom Fields, Support Vector Machine
PDF Full Text Request
Related items