Knowledge Graph Construction of Personal Relationships

Abstract

Knowledge graph has attracted much attention in recent years. It is a high-level natural language processing (NLP) problem, which includes many NLP tasks such as named entity recognition, relation extraction, entity alignment, etc. In this paper, we focus on the entity of persons in the large amount of text data, and then construct the graph of personal relationships. Firstly we investigate how to recognize person names from Chinese text. Secondly, we propose a comprehensive approach including Improved BiGated Recurrent Unit and syntactic analysis to extract the relations between different persons. Then, we align the person entities through entity alignment techniques and some manual proofreading work. Finally, we apply this graph construction process in text records for experimentation. This process performs effectively and efficiently to construct the graph of personal relationships from unstructured Chinese text, and this graph can provide significant relationship insights in texts.