Research On Identification Of Person Name Transliteration

The unknown word recognition is the key technical of automatic segmentation. Transliteration is included in unknown word, Automatic segmentation plays a very important role in information retrieval, information extraction, knowledge discovery and so on.The paper Studies the Chinese Name recognition model and the existing research results of transliteration, then establish the recognition model of transliteration.Firstly, basing on statistical to compute frequency of the word used in the dictionary which named "Handbook of English transliteration"; Secondly, according to the Chinese name segmentation model acquire the original result and computing the reliability and creating the chain of potential name; Thirdly, filtering each potential names by some operate; Then, adjusting the potential names chain; Finally, comparing the original results and the results of recognition of Transliteration and select the optimal solution.In the Improvement, the paper considers the process of filtering and recall process. It optimizes the parameters of frequency and credibility of the names of in some special circumstances.At last, searching a large corpus for testing from the international news in Sina, The result shows that precision ratio reach90%in closed test, but only60%in open test; after the improvement, it can reach70%in the sampling test.