Design And Implementation Of The Relationship Among The Characters Based On Knowledge Graph

The main of police information work is the relationship among the person, organization, and account. Actually, often requires obtained with all relevant information by a person name, for example who he recently contacted, which activities take in, what social account used, these requirements usually need artificial in the mass of information to find the answer. And then, the proposed of this thesis is building a knowledge graph contains the relationship among the characters, thus can get relationship among the characters by query knowledge graph, like the basic information, the related activities, the related characters information. Of course, knowledge graph is used to bring convenience to the police information work, but how to design and build the knowledge graph is difficult.However, much of the extant research work have assumed that the original data have cleaned, character relationship has been built into the triple data, even the knowledge graph has been completed construction, and mainly studies of knowledge graph is in analysis methods and in application scenarios. Therefore, the main work of this thesis is focused on the process from the original data to the formation of the knowledge graph, and as for application, only need to meet the requirements of the relationship of the characters.For design and construction of the relationship among the characters of the knowledge graph. There are mainly three difficult problems: one, the very large amount of original data and data structure is completely different, how to extract to a person, organization, or an account, and how to judge the two characters exist relationship. Two, for the update of knowledge graph, how to judge whether the newly added characters already exist in the knowledge graph, and when the characters already exist, how to merge the two characters. And character relationship contains thousands of kinds of relationship between people and people, people and organizations, people and website, and people and account, how to design each kind of relationship model, which both describe the basic information about the object, and also describe the relationships between objects.The main work of this thesis is:(1) On the basis of ontology modeling, the modeling method of character relation is put forward. Firstly, according to the definition of domain, class, attribute and entity, this thesis designs four kinds of data structures in detail, and guides the creation of the character set and the relation set, and proves the feasibility of the model.(2) Based on the natural language word segmentation technique, the extraction technique of character entity with multi regular expression is proposed. ICTCLAS and LTP are compared through experiments, the different characteristics of the two kinds of segmentation technology are analyzed in this thesis. At the same time, it is proved that the combination of multi regular expressions can improve the effect of entity extraction, and it is especially suitable for the identification of the account entity.(3) This thesis puts forward three kinds of application schemes, which are based on the knowledge graph, semantic search, scene search, get the implementation details of the three application schemes are compared and analyzed.