Building A Chinese Entity Relation Graph Based On Hypernym

With the rapid development of the Internet and big data, how to extractvaluable information to dig out large-scale data, and collate this information toconstruct the system has become an important research topic at home and abroad.For text messages, the established relationships between entities, and further build ahuge wealth of the classification system has far-reaching significance for solvingsearch problems, and many natural language processing problems.Entity is the basic unit of natural language processing, entity-relationship isthe foundation of natural language processing issues, is the basis of many othernatural language processing tasks, but also the basis of many information retrievalproblems. Build an accurate and comprehensive entity relationship map has greatacademic significance and application value.This article focuses on building an excellent performance, large-scale Chineseentity relationship map, where relationships between entities are mainly hyponymy.The main contents of this paper includes three aspects: the relationship between thehost entity recommendation algorithm based supplements, based on mining frequentitem sets word association upper hierarchical and structural information based onthe word hypernym hierarchy.This article uses the frequent item sets mining association with the wordmethod of combining structural information analysis method of combining PC toautomatically dig host relationship between words, to achieve a better level ofeffect, and innovative use of the recommended algorithms hypernym entity to besupplemented. With a high level of accuracy of results, so that a large number ofentities to achieve automatic classification results. For each method, the paper havecarried out a rigorous reasoning and assumptions, design the best solutions forspecific problems. In the set of relevance in the process of mining frequent items,we use the Apriori algorithm and its improved for practical problems. In theanalysis of word structure information in the process, we focused on Chinese wordfeatures unique core design program corresponding algorithms. Experimental results show that the Chinese entity-relationship database in theperformance of this article can be applied to a degree, but with the increase in thenumber of users of the system, the size and quality of the host relations will befurther improved. Also used in the experiments auxiliary efficient optimizationalgorithm and data structure, such that a substantial increase in efficiency of thesystem.