Domain Entity Relationship Extraction Based On Word Vector And Deep Convolutional Neural Networks

With the rapid development of internet technology, the internet has transformed the web of documents into semantic web, which makes the knowledge graph construction become a hot research topic. Especially for the specific domain, knowledge graph is the foundation to realize the personalized service, and building the knowledge graph of the specific domain has become the focus of the research. Therefore, how to semi automatic build a domain knowledge graph is a very meaningful work. This paper has carried out the following research on the schema system construction and the relation extraction task in the domain knowledge graph construction.1 Hierarchical relation extraction of domain entity. Hierarchical relation of domain entity is the skeleton of knowledge graph, which determines the depth of knowledge graph. In this paper, a method based on bootstrapping and word embeddings is proposed to automatically extract the hierarchical relationships of the domain entities. First, The similarity calculation method based on word embeddings is used to cluster the patterns; Screening the mode, which have high confidence, to identify relationship of unlabeled corpus; Then, the bootstrapping method is used to iterative extract the relation instance. Finally, according to the offset of the word embeddings and combining entity hierarchy field characteristics, using the linear mapping learning method to construct the semantic hierarchies structure.2 Semantic relation extraction of domain entity. The semantic relation of the domain entities is an important part of the domain knowledge graph, which determines the breadth of knowledge graph. In this paper, we propose a method based on the convolution neural network to extract the semantic relations of the domain entities. The convolution neural network is used to automatically learn the lexical features, context features and the sentence level features; Then, training entities semantic relationships classification model used this feature. 3 Drawing domain knowledge graph. The knowledge graph is a complex model, which the nodes represent entities (Concepts), and the edge represents a complex network of semantic relations between entities. Therefore, traditional relational database is difficult to express these complex relationships. In this paper, we propose a method based on Neo4j to draw the domain knowledge graph. Representing relationship of the domain entities by the graph database, and visualization display the domain knowledge.