面向流域水资源自由文本的属性抽取方法

Attribute Extraction from Free Text on Basin Water Resources

[目的/意义]为流域水资源领域知识库的构建提供数据来源。[方法/过程]针对非结构化的流域水资源属性信息,提出一种基于属性触发词的流域水资源属性抽取方法。首先,基于统计学方法分析流域水资源文本,得到流域水资源实体-属性触发词-属性值的分布规律;其次,利用频繁模式挖掘,提取出属性触发词;最后,结合属性触发词与属性触发规则,实现属性三元组的抽取。[结果/结论]经百度百科自由文本实验与对比分析,该方法适用于数值型属性抽取,具有较高的精确率和召回率。

[Purpose/significance]The paper is to provide data sources for repository construction on basin water resources.[Method/process]According to unstructured attribute information on basin water resources, the paper puts forwards an attribute extraction method based on attribute trigger words. Firstly, the text of basin water resources is analyzed based on the statistical analysis,the distribution rule of entity-attribute trigger word-attribute value is obtained; Secondly, frequent pattern mining is used to extract attribute trigger words; Thirdly, triple attribute extraction is carried out according to attribute trigger words and trigger rules. [Result/conclusion]Through experiment of Baidu Encyclopedia free text and comparative analysis, the results show that the method applies to numeric attribute extraction, and the precision and recall are quite high.