Word Segmentation and Named Entity Recognition for SIGHAN Bakeoff3
Suxiang Zhang, Ying Hua Qin, Juan Wen, Xiaojie Wang · Meeting of the Association for Computational Linguistics · 2006
We have participated in three open tracks of Chinese word segmentation and named entity recognition tasks of SIGHAN Bakeoff3. We take a probabilistic feature based Maximum Entropy (ME) model as our basic frame to combine multiple sources of knowledge. Our named entity recognizer achieved the highest F measure for MSRA, and word segmenter achieved the medium F measure for MSRA. We find effective combining of the external multi-knowledge is crucial to improve performance of word segmentation and named entity recognition.