The service of recommending literature has been one important content of knowledge services. This paperintroduces the technology of paring textual orientation in Natural Language Processing. After parsing the semantic structure of the sentences in citation text, the semantic structures are transformed into the 2-gram or 3-gram model for judging the orientation of text. Therefore, the subjective information about references in the citation text is obtained. In addition to being cited, the reference itself also review other references. A recommendation-index is calculated by combining the parsing result of the orientation in citation and its review, to nominate the worth reading literature in a set of papers. The experiment in this paper shows that the method of recommending literature can be applied with a decent precision. However, there are some development to expand the word knowledge base, and improve the performance of parsing the semantic structures of sentences and analyzing the semantic relations between words.%文献推荐服务已经成为数字图书馆的重要知识服务内容之一.本文引入自然语言处理中的文本倾向性分析技术,通过对引证文本的语句语义分析,把语句的语义结构转化为倾向性分析的二元或三元模型,得到引证文本对参考文献的主观评价信息.结合参考文献本身的正文中对其他文献的评论指数,给出了文献推荐度的计算方法,从而实现对文献集中文献的自动分析和推荐服务.实验表明基于文本倾向性分析的文献推荐服务是可以实现的,并具有较高的准确率.在实际应用中还需要扩大词语知识库的规模,并提高语句的语义结构分析、词语语义关系分析等方面的性能.
展开▼
