基于多维语义的互联网药品信息提取方法

摘要

A multidimensional-semantics based Web information extraction method is proposed in this article to extract medicine information on the Web. The method overcomes the heterogeneity of Web pages from different sources and finds the common characteristics among them by building up a semantic dictionary and describes the knowledge of medicine information over the Web. At the same time, it utilizes a structural-semantic-entropy-based approach to detect data-rich sections on Web pages, then extract information of interest from them and finally verify and supplement the extracted information by generating extraction rules using Xpath. The method is able to obtain information from heterogeneous sources both automatically and effectively. Experiments shown that it has high precision and recall, thus can provide sufficient information for the government to enhance supervision of medicine market on the Web.%提出了基于多维语义的互联网药品信息提取方法,构建语义词典通过从多个维度对互联网药品知识进行描述,克服了不同来源网页之间的异构性并找出了其隐藏的共性.同时,采用了基于结构语义熵的方法对目标网页信息聚集区域进行定位,从中提取感兴趣的药品信息.最后再通过语义词典对提取的信息进行验证并自动生成XPath提取规则进行补充.该方法能够自动有效地从互联网的多个信息来源获取药品信息,实验证明其具有较高的准确性与召回率,可以为政府相关部门加强互联网药品市场监管提供足够的信息依据.