基于引文和标签的论文推荐算法研究

Research on paper recommendation algorithm based on citations and labels

设计一个共联与标签相结合的论文推荐模型.首先,提出在传统共联算法上根据引文被引用次数(引文热度)赋权,减小热门引文对计算论文间相关性的影响程度.其次,建立多级分类标签对论文进行层次分类,建立论文在研究主题和行文结构上的分类树,并提出了父级和跨度的概念,从论文研究课题的深度和广度方面讨论两篇论文之间的相关性,给用户提供高质量的推荐服务.

We propose a paper recommendation model which combines co-coupling algorithm and label strategy. Firstly, we weight a paper by its citation frequency (heat of citation) based on the traditional cocoupling algorithm, to reduce the influence of popular citations on the relevance calculation among papers. Secondly, we propose using multilevel labels to classify papers hierarchically, and build a classification system concerning the topic and structure of papers. Meanwhile the concepts of parent level and level span are proposed, which measures the relevance between two papers from depth and breadth of the paper topic. This model can make recommendation services better.