Research And Implementation On The Method Of Chinese Domain Concept And Relation Extraction Based On Semantic Graph

With the rapid development of Internet technology,the amount of data is increasing rapidly,it becomes more and more important to mine effective information from massive data.Nowadays,in the field of knowledge engineering,knowledge graph,especially domain knowledge graph,plays an important role and becomes the infrastructure of Internet knowledge-driven intelligent applications.For the construction of domain knowledge graph,the data schema of knowledge graph should be constructed first.Because the domain data has a huge size and is often constructed by unstructured texts,how to automatically construct the data schema of knowledge graph becomes the key issue in this research field.The extraction of domain terms,domain concepts,and relationships are important sub-tasks in constructing the data schema of knowledge graph.The main contents of this paper are as following.1)A method of Chinese domain term extraction based on hybrid strategy is applied.Firstly,lexical analysis of domain data sets is performed,and the candidate domain terms is extracted based on rules.Secondly,candidate domain terms are filtered by using statistical methods.The TF-IDF(term frequency–inverse document frequency)algorithm is used to extract single word of domain terms,and the TextRank algorithm is used to extract single word of domain terms and multi-word terms.Experimental results show that the method makes the extraction of domain terms more comprehensive.2)A Semantic Graph Based Chinese Domain Concept Extraction(SGCCE)method is presented.Semantic graph of terms that containing semantic information is constructed firstly.Then community detection algorithm is used to analyze and divide the term semantic graph to extract domain concepts.Experimental results show that the full use of semantic information can get better results of concept extraction.3)A framework of concept relationship extraction based on semantic features is proposed in this paper.First,Semantic Graph Based Non-taxonomic Relationships Identification(SGNRI)method is used to identify the relationship between concepts.Then,Dependency Parse Based Non-taxonomic Relationships Labeling(DPNRL)method is used to label the relationship between concepts.The experimental results show that the best results which is obtained by fully integrating syntactic and semantic information.