基于贝叶斯分类的中文物种描述文本的语义标注研究

Semantic Annotation of Species Description Text in Chinese Literature by Naive Bayes Classifier

本研究从《中国植物志》中随机采集1000个文档作为数据集，采用基于先导词的朴素贝叶斯算法实现中文物种描述文本的自动语义标注。通过实验性研究，实验数据表明，先导词能够有效提升朴素贝叶斯的标注效率。采用先导词后，F平均值提高0．048～0．107，尤以Fr为2时效果最好，整体标注性能F平均值高达0．902。各元素的标注性能也较为理想。Fr分别取1、2、3时，大部分元素的F值为0．730～0．964。

Based on leading words, this paper using NaYve Bayes to implement the automatic semantic annotation of species description text in Chinese with the data set of 1000 document which coliected from Flora of China randomly. Through experimental research, the data indicates that the leading words could effectively enhance the markup efficiency of NaYve Bayes. With the leading words, the average values of F increase 0. 048 to 0. 107. Particularly, when Fr takes the value of 2, the overall markup performance best with the average value of F achieved 0. 902. Markup performance is also ideal for each element. When Fr takes the value of 1 , 2, 3 respectively, the value of F of most elements is between 0. 730 and 0. 964.