科技文献跨语言推荐模型研究

摘要

Information overload and language barrier seriously affect the efficiency of acquiring academic literatures in foreign language and how to help users obtain their targeted literatures becomes an urgent problem for digital libraries. It is reported that personalized recommendation systems can well deal with information overload problem, but most of the current researches are based on single language and seldom discusses the methods of recommending academic literatures under a multilingual environment. The paper proposes a framework of cross-language recommendation system of academic articles, and its relevant modules are described in details, including extraction of users＇ interest feature, language translation and hybrid recommendation etc. This research is expected to expand the research content of personalized recommendation systems and provide technical solutions for improving the efficiency of obtaining academic articles in other languages. 3 figs. 18 refs.%信息超载和语言障碍影响我国科研人员对外文科技文献的有效获取，如何提高获取效率成为亟待解决的问题。个性化推荐能很好地处理信息超载现象，但当前国内外相关研究都基于单一语种进行，多语种环境下的推荐研究非常缺乏。本文提出网络环境和海量数据背景下的科技文献跨语言推荐模型，并论证用户兴趣特征抽取、语言转换和混合推荐等步骤。利用Web日志挖掘技术，分析基于多种信息行为的整合分析方法抽取用户兴趣特征，以分类表作为参考体系建立用户兴趣表示模型，在用户一特征词转化为用户一类目矩阵的基础上开展推荐研究。图3。参考文献17。