Research On The Application Of Linguistic Features In Chinese Semantic Relation Extraction Between Named Entities

Semantic relation extraction is a critical subtask in Information Extraction, which aims to extract relationships between name entities from natural language text documents.This paper adopts an approach based on basic linguistic features to address the semantic relation extraction problem. First, a series of basic linguistic features are devised, includeing words and their context, entity type, entity mention type, entity overlap type, and basic phrase chunks. Then, we use combine these basic features, leading to combination features, such as word combination, entity type combination, entity mention type combination, and basic phrase chunk combination etc. Finally, a feature-based SVM classifier is applied to extracting semantic relationships.Experimental results on the ACE RDC 2005 Chinese benchmark corpus show that, the F-measure of relation detection and major type relation classification based on our method are 60.81 and 56.64 respectively. This suggests that linguistic features achieve reasonable results in Chinese semantic relation extraction. Furthermore, this paper also investigates the contributions of different featurers to relation extraction, which demonstrates that the word feature and entity type feature contribute much to Chinese semantic relation extraction.