A Novel Method of Automobiles' Chinese Nickname Recognition

摘要

Nowadays, we have noticed that the free writing style becomes more and more popular. People tend to use nicknames to replace the original names. However, the traditional named entity recognition does not perform well on the nickname recognition problem. Thus, we chose the automobile domain and accomplished a whole process of Chinese automobiles' nickname recognition. This paper discusses a new method to tackle the problem of automobile's nickname recognition in Chinese text. First we have given the nicknames a typical definition. Then we have used methods of machine learning to acquire the probabilities of transition and emission based on our training set. Finally the nicknames are identified through maximum matching on the optimal state sequence. The result revealed that our method can achieve competitive performance in nickname recognition. We got precision 95.2%; recall 91.5% and F-measure 0.9331 on our passages test set. The method will contribute to build a database of nicknames, and could be used in data mining and search engines on automobile domain, etc.