Chinese Named Entity Recognition with new contextual features

Chinese named entity recognition (NER) is studied in two directions: inner structure and outer surroundings. Inner structural analyses induce constitutions of person, location and organization name from the point of linguistics. However inner structural rules for named entities only provide necessary conditions for a sequence of Chinese characters being an entity name but not sufficient. Whether a string being a proper name or not is also determined by its contextual information or sometimes common sense. We build Chinese NER system based on supervised machine learning using features induced from simple inner structure and contextual information. We compare some NER approaches. The experimental results indicate complicated cases of various NER strategies. Then this paper turns to explore contextual features of named entities on large scale corpus, seeking for contextual evidence for different strategies of NER and mark words giving clues to the occurrence of NE. Finally, we apply some conclusions to improve NER system by enriching features in model and enhance the performance distinctly.