Relationship Extraction Tactics of Chinese Entity Based on Formal Concept Connectivity Distance

As Chinese expression diversity, there are some shortcomings in traditional algorithms of Chinese entity relationship extraction. For example, workload of labeling by hand on training corpus is too large, the generated relationship schemas usually have poor versatility, and it is difficult to select or integrate high quality domain ontology for extraction task. Moreover, these algorithms don't consider the fact that the entity relationship usually has different meanings with the different topic backgrounds or with the various concept granularities. The paper, utilizing statistical method and linguistics knowledge, carries out the work of crawling, parsing, filling, builds the relational formal concept lattice with Chinese entities context, and acquires entity relationship schemas described by relational formal concept. With these relational schemas and concept built above, we carry out the entry concept correlation computing and the predicate text flexible matching, and get the concept connectivity distance between entities to achieve the non-single and indirect entity relation extraction. The granularities of concept in relation extraction are more flexible, and the relational schema described by formal concept is more versatile and robust. The method in this paper provides a better semantic description for the extracted relationship, and obtains a good relation extraction performance.