Automatic Recognition Of Relation Words In Chinese Complex Sentence Based On Dependency Grammar

At present, Chinese information processing in the two aspects of word processing and word processing have achieved good results. Sentence processing as the next target of word processing, and will be a bridge to better transition to discourse processing, should be particularly pay attention to the research. Complex sentences accounts for a large proportion in Chinese. Research of Complex Sentences has important significance in Chinese information processing. Current research on complex sentences are start from marked complex sentences. Relationship flag in complex sentences is the hub of the various clauses. Relation words in the sentence to the semantic and hierarchy plays an important role. Therefore, the study of relation words in complex sentence is very necessity.First, the existing software and complex sentences relations word ontology, given word tagging to relation word in the selected sayings compound whether it acted as a quasi-relationship word. Combined with manual annotation approach, the part of the corpus has been amended. Complex sentences that have been marked to extract features. And These complex sentences use dependency grammar analysis platform, obtained dependency syntax tree. Then extract dependency features from dependency grammar tree. And combine the two aspects characteristic of their own feature and dependency feature, represented as feature vectors, structure training and test sets. Using the training set, SVM train a model, then use the trained model to predict the test set.In order to test the feasibility of this method, select sets of typical relation words from the complex sentence corpus.to have recognition for these relation word phrases. Experiment has achieved good results, automatic identification of relation words have a higher accuracy.This suggests that the research for relation word in complex sentences based on dependency syntax is feasible.