基于词典匹配的蒙古文命名实体识别研究

A Study of Mongolian Named Entity Recognition Based on Dictionary Matching Method

命名实体是自然语言处理中重要的信息元素,是正确理解文本的基础。本研究根据蒙古文命名实体的特点,采用机器词典进行命名实体的最大化匹配后引入有限状态自动机。实验表明,采用该方法对10万词级蒙古文新闻语料库进行标注后,其召回率为90.67%,准确率为92.53%,F值为91.59%,取得了较好的识别效果。

Named entities are important information elements in natural language processing and are the basis for correct understanding of text. Based on the characteristics of Mongolian named entity, this paper introduces the finite state automata with the maximization of named entities by machine dictionary. The experimental results show that the recall rate is 90. 67% , the precision is 92. 53% , the F-measure is 91. 59% , and the recognition effect is achieved on 100, 000 words of Mongolian news corpus.