A Study On Boundary Recognition Of Modern Chinese Prepositional Phrase

Right boundary ambiguity of Chinese prepositional phrase is one of the most important phenomena in Chinese structure ambiguities, which made Chinese syntax analyzing very difficult. The goal of this paper is to recognize the boundaries of the prepositional phrase automatically without using other more complicated syntax analyzing. We hope it can supply some help to the syntax analyzing as one of the parts of syntax analyzing preprocessing.We mainly used statistical method in our study, estimated the maximum likelihood probability of every word behind the preposition to be the right boundary. According to the parameter estimating, we assume that the prepositional phrase right boundary is only related to the type and pos of the two words before and behind the boundary mark. Although this model is not very accurate, it depends on the two neighbor words occurrence probability, we still have to solve sparse data because of our limited training corpus. To get more reliable probability, we used deleted interpolation to smooth the parameters. At last we applied three manual rules to modify our results.We did experiments on "在/pzai" prepositional phrases, The highest accuracy for the closing test and opening test are 97% and 93% respectively. Comparing to the former study, the accuracy has been improved, and solved some former problems.Wang Lixia (Linguistics and Applied Linguistics) Directed by Dr. Sun Honglin...