A Method Of Domain Compound Concept Extraction Based On Multilevel Filter

The domain ontology is a hot topic in artificial intelligence filed,and the domain concept is the basic part of domain ontology,so the identification and extraction in domain compound concept is a basic research work.With the social progress,the development of science and technology,new concepts emerge in an endless stream,especially for the compound concept in each field.These domain compound concepts generally are noun phrases which are formed by domain atomic concepts or words,they refer to more precise information for domain concepts.Identification and extraction in domain compound concept is the basis of the domain text information processing,it has important significance on domain ontology's construction and application,text information retrieval and text mining.The existing word segmentation system can't recognize the new domain compound concept,so it can't meet the needs of practical applications.Therefore,automatic extraction in domain compound concept is needed.This paper builds a multilevel filter extraction model by fusing the thought of statistics and language rule.Firstly,the extraction model screening out domain atomic concept set by using method of improved TF-IDF.We secondly build a space combination rule,screening out initial domain compound concept set.Ultimately we screening out finally domain compound concept set by using POS rules template matching via POS analysis.This paper constructs a verification system for domain compound concept extraction which is based on multilevel filter.We have made experiment on it,and calculated accuracy rate?recall rate and F value.At the same time,we used other two methods to do the experiment.By comparing the result of experiment,we can find that,the accuracy rate?recall rate and F value which calculated by our method is higher than other two methods.So?our method is better than other two methods in domain compound concept extraction.