Automatic Recognition Of Chinese Noun Phrase Based On Probabilistic Context-free Grammar

Natural Language Processing (NLP) is concerned with the automated computerunderstanding of human language. Noun Phrase (NP) recognition is one of the most criticalcomponents of this task, and it is directly related to the text analysis and the correctness oftext processing. In the task of the information retrieval, the noun phrases are the mainrecognition target. The paper analyses the structure of the noun phrase, and then recognizesthe noun phrase in the sentence using the merging method which combines the statisticalmethod and the parsing method. This paper mainly contains the following points.First, this paper uses the Probabilistic Context-Free Grammar (PCFG) recognizes thenoun phrase. In order to reduce the three hypothesis of the PCFG, this paper uses the PCFGwhich combines the information of the context, splits and merges the nodes, and uses the“coarse to fine” to analyze the noun phrase.Second, this paper proposes a recognition method which is on the basis of AuxiliaryPhrase Mark. Based on the detailed analysis of the phrases of different classification systems,this paper gives a mapping formula. According to the formula, there is a mapping between thephrases of different classification systems. Then, based on the mapping result and the probabilitydistribution of phrases, the auxiliary phrase marks are combined. Experimental results show thatthis method effectively shortens the recognition time of noun phrase without reducing the F-value.Third, by utilizing the advantages of the statistical method and the parsing method, thispaper proposes a noun phrase recognition method on the basis of merging the results of theConditional Random Field (CRF) and the Probabilistic Context-Free Grammar (PCFG).Experiment results show that CRF has a better effect on underlying noun phrase recognitionthan PCFG, while PCFG has a better effect on upper noun phrase recognition than CRF.Therefore, this paper combines the recognition results of CRF and PCFG to achievecomplementary advantages and improve the recognition precision of noun phrases.Through the above analysis, we confirm the method for the noun phrase recognition, and do the experiment for the noun phrase. The results show that the methods are effective.Future work is required for obtaining better performance.