基于主题模型的跨学科协作文献推荐

摘要

Web中存在着海量的各类科技文献,研究人员虽然可以利用各种搜索工具对这些文献进行检索,但是,如何高效地找到与自身研究相关的文献变得越来越困难.最近出现的一系列在线研究者社区为解决这一问题提供了一种新的方案.提出一个基于主题模型的协作文献推荐,此模型将传统的协同过滤和概率主题模型,以及知识协作网络模型相结合,提供了一个可判别的隐语义结构.在考虑不同的用户评价所给出的文献索引率,以及新发表的文献的主题分布的基础上,利用语义相似度的计算工具,提出基于概率的跨学科的检索推荐.采用来自于CiteULike的一组数据,验证了该方法的有效性和可行性.%There exists plenty of all kinds of literature, although the researchers can use some sorts of searching tools for literature retrieval , but how to seek out more efficiently relevant literature becomes more and more difficult Recently appearing in a series of online community for the researchers is a new solution. A model based on topic model of interdisciplinary literature recommendation was presented, the combination of traditional collaborative filter and probability topic model, and knowledge collaboration network model has been put forward, so it provides a latent semantic structure which can be distinguishable. In terms of different user's appraisal of the given literature index rate,and topic distribution of newly published literature in the foundation,This paper used semantic similarity calculation tools,and puts forward the retrieval recommendation for interdisciplinary research based on probability. For this reason, we studied a set of data from the CiteULike,and experimental results show the feasibility and effectiveness of this method.