Esearch On Chinese Person Disambiguation Based On Sentential Semantic Structure And Personal Implicit Relationship

With the advent of the Internet plus era, the Internet contains rich personal information.But the phenomenon of the single name for several persona, personal information scattered and Chinese text structure diversity results that information retrieval results has been difficult to meet the needs of the people. The process of information retrieval is mainly the personal names disambiguation that in Natural Language Processing has a great application value. It is the basis of data mining, people’s social computing, and construction of the resource information database of the people. A wide range of applications in the field of automatic question answering system, emotion analysis, and the strength analysis of the relationship between people and so on. Name ambiguity is a special case of identity uncertainty, and the identification of a person’s identity depends on the individual’s information, so how to make better use of the existing information in the text is the key to the research of personal name disambiguation.Many existing methods of name disambiguation based on human features are not suitable for documents due to their sparsity. For solving this problem, we propose a human social relationship based Chinese name disambiguation by analyzing sentential semantic structure. Firstly, we build a graph of social relationships according to the characters of human social relationships. Then, we cluster the relationships by combing with the attributes of the name entity, such as his career and unit, as auxiliary characters. After name disambiguation, original documents can be summarized by the name which the documents discuss. The experimental results show that the analysis of sentential semantic structure is able to improve the precision of personal relationships and attributes,so it effectively improves the performance of name disambiguation.The extraction of the direct relationship of the person is to be built in the sentence, which must include relation characters that it has a significant dependence with personal name.When the sentence has no obvious relation characters to indicate between the name entities of social relations, but according to the semantic analysis, there is a social relation between name entities. For solving this problem, this paper extracts implicit relationship between name entities by combing the analysis of the sentence structure with the statistical method. In the first place, we analyze the dependence relation between the semantic components of the character entity in a sentence according to the sentential semantic structure, and infer the relationship type by the transfer characteristic of some relation characters. If there are many kinds of transfer types for a character, and it is difficult to infer the relationship type.According to the transitivity and exophilicity of relationship between entities, on the basis of the constructed relation network graph, the relationship strength between the nodes is calculated based on statistical method, and the type of relationship will be determined by setting the appropriate threshold. At last, the relationships extracted of two methods are integrated into the network to further improve it. According to the experimental results, the proposed method can effectively improve the recall rate of Chinese disambiguation.