Title: A Recognition Approach Study on Chinese Field Term Based Mutual Information /Conditional Random Fields
Abstract: A new auto-recognition approach based on mutual information/ Conditional Random Fields (CRFs) was put forward in this study. Firstly, statistics-based mutual information algorithm was applied to separate the Chinese words accurately, then the sub-words were picked out from the accurate separation according to the entropy of the left and right information. Secondly, the relative frequency of the sub-words was calculated. Thirdly, three training characteristics, including words, part of speech and relative frequency, were used as training datasets to obtain a model for field terms characters by CRFs. Thirdly, the Chinese words recognition was accomplished by the CRFs model. Finally, a practical experiment was executed and the results showed that the precision, percentage and Fmeasure of the recognition is 78.63%, 87.10% and 82.65% respectively, which is significant better the normal mutual information/ Conditional Random Fields (CRFs) algorithm.