Recognition Of The Chinese Name Based On Maximum Entropy Model

Automatic recognition of Chinese name is one of the most difficult problems in word segmentation, and because of this bottleneck; the precision of word segmentation is not high. Automatic recognition of Chinese name is a important part of many information process fields such as QA (Question and Answer), IE (Information Extract), IR (Information Retrieval). How to distinguish and classify correctly named entity is a key and difficult work.Due to two reasons Chinese name recognition is much more difficult than that of English. The first is that there is no space between words in a sentence, causing low the precision of word segmentation. The second is that there are not obvious features of accidence for Chinese to identify the names (such as capitalization of places or human names), which makes it difficult to recognize names.By presenting an approach For Chinese name recognition based on maximum entropy, this paper breaks through the traditional recognition method based on the rules and statistics and gives a solution method for Chinese name. The method uses the advantage of maximum entropy model, also avoid the difficulty in features select caused by the large difference in framework between different name entities.Firstly, this paper analyses the structure of Chinese names and their complex appearance in texts. At the same time, taking the feature of Chinese names into account and considering context information, word information, dictionary information, label information and so on, the author puts forward the feature templates adapted to Chinese name recognition, and gets the candidate feature space by combining atom feature templates, then puts forward the arithmetic of feature adopting...