A study of Chinese named entity and relation identification in a specific domain

This thesis aims at investigating automatic identification of Chinese named entities (NEs) and their relations (NERs) in a specific domain. We have proposed a three-stage pipeline computational model for the error correction of word segmentation and POS tagging, NE recognition and NER identification. In this model, an error repair module utilizing machine learning techniques is developed in the first stage. At the second stage, a new algorithm that can automatically construct Finite State Cascades (FSC) from given sets of rules is designed. As a supplement, the recognition strategy without NE trigger words can identify the special linguistic phenomena. In the third stage, a novel approach positive and negative case-based learning and identification (PNCBL&I) is implemented. It pursues the improvement of the identification performance for NERs through simultaneously learning two opposite cases and automatically selecting effective multi-level linguistic features for NERs and non-NERs. Further, two other strategies, resolving relation conflicts and inferring missing relations, are also integrated in the identification procedure.