专利领域同义词识别

Recognition of Synonyms in Patent Domain

专利领域的同义词表是提升专利检索性能的重要资源,当前国内此类资源相对稀缺,亟需寻找一种高效的同义词自动识别方法.该文提出一种用于专利领域同义词识别的类决策树模型.分析专利搜索日志中候选同义词构词规律、外来词音译规律、汉语语义词序规律,提出词形特征、语音特征、语义特征.分析特征间的关联特性以及样本数据的分布规律,构建用于出专利领域同义词识别的类决策树模型.实验结果表明,在多个测试样本中,平均F-值达到0.914,取得了较好的识别效果.

The synonym table is a necessary resource for enhancing the performance of a patent search engine. It is urgent to find an ef- ficient automatic recognition method for synonyms in patent domain. This paper presents a Likely Decision Tree model for the recogni- tion of synonyms in patent domain. By analyzing the formation rules of candidates,the translation rules of exotics and the word order rules of Chinese semantics, literal feature, pronunciation feature and sematic feature are presented. A Likely Decision Tree model is presented to, by analyzing the association between those features and the distribution of the sample data. The experimental result shows that this method can achieve better recognition results with an average F-value of 0.914 in test samples.