Automatic Chinese Sentences Group Method Based On Coreference Resolution

Automatic Chinese sentence grouping is to divide a text into text fragments with different themes,and it plays a very important role in information extraction,abstract generation,chapter comprehension and so on.Name disambiguation and different names of a same entity aggregation is not only a key to coreference resolution,but also the focus and a difficult problem of natural language processing.Currently,the focus of automatic Chinese sentences grouping work is to recognize the boundaries of different topics instead of using the coreference relations of a passage for language comprehension,or for the reason that inaccurate results were usually existed due to vagueness resolution.Thus,the key technologies are described in detail by reading a lot of theories and do the same to the methods of sentence group and coreference resolution.Then,the characteristics of the existing methods for resolution are deeply analyzed.It is proposed to add the cluster information of the entity to the existing model and replace the simple distance-first principle with the optimal strategy of combining the distance and the frequency attribute.The features property and optimization strategy was validated through the three species mainstream methods of coreference resolution.The experimental results show that the optimal strategy in this paper can efficiently complete the digestion task by combining the cluster information of the entity.The contribution rates of a single feature was gotten through the method of coreference resolution with level filter model.So the features made big contribution to coreference resolution could be gotten,which are matching the exact string,alias,appositive.Finally,Automatic Chinese sentence grouping based on coreference resolution was proposed based on the previous research work.The features making effective contribution to coreference resolution were chosen in the method.Then the referring information will be obtained based on multi-layer filter of Chinese coreference resolution,which will solove the problems of different entities refer to the same entity or entities refer to unknown entities.A set of evaluation functions were designed based on MDA combined referring information and the cohesive function of pronouns in sentence combinatio,Lastly,the value of increased by about 7%,so automatic sentences grouping was improved with the method propounded in this paper.