Utilizing the Relation Sets of Entity Pairs to Recognize the Organization Names in Chinese Short Text

This paper proposes a method of recognizing the organization names in Chinese short text by utilizing the relation set of entity pairs according to entity relation theory. A three-step strategy is involved in the main idea of this paper: (i) Extracting the relation sets of entity pairs by using the inherent structural characteristics of Wikipedia. (ii) Vectorizing the context of relation sets of entity pairs and constructing the vector space model. (iii) Finally recognizing the organization name by similarity comparison combining with the method of using the HMM to recognize named entity and the method of using the search engine to reconstruct corpus. Through the experiment, the recall in this method reaches 58.72% and F ₁ reaches 67.48%, which increase by 13.83% and 9.28% respectively compared with the results got by using Hidden Markov Model (HMM). Proved by the experiment this method is effective.