Chinese Relation Extraction Using Web Features and HNC Theory

摘要

Chinese named-entity relation extraction is a key step in the task of Chinese information extraction. Feature-based method is one of the main methods of Chinese relation extraction. In this new method, Web co-occurrence feature and Bag-of-word (BoW) correlation feature are introduced, and the words similarity is defined based on HNC theory. Experimental results showed that the F-score was improved by this method, and both of the features are effective on Chinese relation extraction.