An Automatic Evaluation Method for Chinese Name Recognition

Abstract

In order to evaluate the methods of Chinese name recognition effectively, this paper presents an automatic evaluation method for measuring the performance of Chinese name recognition systems by using the idea of randomly selecting sentences and replacing names. The method can be mainly divided into three parts: the construction of evaluation corpus, the generation of evaluation files and the computation of evaluation scores. As evaluation files are generated automatically and randomly in the method, the method can overcome the disadvantages that traditional standard evaluation files usually have such as small-scale, fixed and un-reusable. Experiments show that the method is not only efficient, valid and objective, but also can be extended to evaluating other kinds of named entity recognition such as place names and organization names.