Tuple Refinement Method Based on Relationship Keyword Extension

Entity relation extraction is mainly focused on researching extraction approaches and improving precision of the extraction results. Although many efforts have been made on this field, there still exist some problems. In order to improve the performance of extracting entity relation, we propose a tuple refinement method based on relationship keyword extension. Firstly, we utilize the diversity of relationships to extend relationship keywords, and then, use the redundancy of network information to extract the second entity based on the principle of proximity and the predefined entity type. Under open web environment, we take four relationships in the experiments and adopt bootstrapping algorithm to acquire the initial tuple set. Three tuple refinement methods are compared: refinement method with threshold set, refinement method with relation extension and refinement method without relation extension. The average F-scores of the experimental results show the proposed method can effectively improve the performance of entity relation extraction.