Automatic Recognition And Rule Mining Of Chinese Relation Words In Compound Sentences Based On Dependencies

Word and words processing of Chinese information processing is already relatively mature,sentence and discourse processing is the difficulty and key point of Chinese information processing to research.Compound sentence which is made up of two or more clauses is the transition of sentence processing to discourse processing,so the compound sentence is an important topic in the research of Chinese information processing.The relation word also called relation mark which used to connect clauses is the main research content in this article,which can fully embodies the logical semantic relationship between clauses,and are also the important signs of clauses structure.Current research about the relationship of word collocation is mainly concentrated in literal features,the syntax analysis was not discussed.Based on the constraint conditions of Chinese compound sentence relational words' collocation which is on account of literal characteristics,this paper studies the constraint conditions research which are based on the dependency relation,using five axioms of dependency grammar for analyzing relation words' dependency relation on Chinese compound sentence,and suumarizes seven kinds of constraints,then formalizes these constraints into dependency relation rule to join the rule base,using rule which combines literal featires with grammar features to automatic identify Chinese compound sentence's relational word.The experimental results show that the rate of the relational words'recognition is above ninety percent.The relation word reeognition rule base has 734 rules,which mainly based on the characteristics of literal,and the rules based on dependencies still need to supplement.This article on the basis of dependency grammar,using FP-tree algorithm of mining frequent itemsets for automatic mining dependency rules in compound sentence.First to pretreatment of corpus,classify the sentence according to the relation words for avoid repeating scan database,to ensure the quality of mining rules remove the small classification results at the same time,then through characteristic analyzer analysis and formal results after get the dependence of characteristic collection,and then using FP-tree algorithm to mining experiment corpus,which totally inining 84 rules.This article research the dependence relationship of relation words' collocation and the automatic mining of dependency rules of Chinese compound sentence from the perspective of dependency grammar.Results based on dependency analysis of compound sentence,and automatic identify relation words by using the dependence relationship of relational words' collocation,then introduce the FP-tree algorithm to automatic mining the potential dependence relationship of relation words,collocation.The experimental results show that the automatic recognition rate of Chinese compound sentence' relation words is improved obviously,and using FP-tree algorithm successful mining 84 dependency rules,which show that this algorithm is feasible and effective in automatic mining dependency rule.