Recognition Of Chinese Prepositional Phrase

Prepositional phrase is a complex structure, a large proportion of Chinese sentences. Prepositional phrase recognition has always been the difficulty of natural language processing problems. The prepositional phrases identified correctly can simplify the sentence structure and it is of great significance in deep syntactic analysis.The purpose of this paper is the segmentation and speech tagging based on the complete identification of Chinese sentences prepositional phrases, for the full syntactic analysis, and other basic natural language processing tasks. In this paper, Conditional Random Fields model for recognition of Chinese sentences prepositional phrases, Distributed based on CRFs distributed strategy and the combination of two-tier CRFs prepositional phrase recognition.Prepositional phrases in sentences in the form of a very flexible, using the one model difficult to achieve the desired results. In order to overcome one model the limitations of various types of prepositional phrases, this paper presents based on CRFs distributed strategy. It would consist of 59 different prepositions in the prepositional form, grouped by type of pre-positions, prepositional phrases of different types of groups select the appropriate type of recognition of the group of sensitive features, combined with different sets of prepositional phrases CRFs build recognition model. Recognition model to identify different result in the consolidation process in the conflict, we use dual-layer CRFs to solve. Dual-layer CRFs can well identify nested prepositional phrases, and the results combined in the conflict are primarily in nested prepositional phrases, so there is a conflict on the part of the two-tier CRFs identified as the final result Results. Finally, the syntax of prepositional phrase rules is used to post-process.Experimental results show that based on a combination of CRFs distributed strategy and two-tier CRFs identify prepositional phrases obtained good results. Open test system achieved 91.46% precision and recall rate of 91.57%, F value reached 91.52%.