TR-SLT-0070 :2004.03.31

Rui Yang, Hirofumi Yamamoto

Unsupervised Chinese Word Segmentation

Abstract:This report presents my work in ATR during the last one and a half months. We tried to use an unsupervised method base on EM algorithm to get a segmentation result with low total entropy. We haven't got the completed result of the experiment till now, but from the trend line, we can say it is effective. But still we have much work to do to reduce the high complexity.