摘要：Link weighting is crucial for performance of link-based web page ranking algorithms. Typically, a link is viewed as recommendation between pages, which is an unquantifiable term and existing approaches lack physical interpretations. In this paper, we view a link as the attractive force between pages, and map concepts of web (in/out degree, content similarity, etc.) to those of physics (mass, distance, attractive force). Inspired by Reilly’s Law of Retail Gravitation, we propose a gravitation-like model for calculating the attractive force. We then implement a instance of our algorithm framework by taking some features of web pages into consideration. Experimental results show that this instance outperforms other typical algorithms (HITS, Randomized-HITS, and SALSA) with higher precision, better resistibility of TKC effect and no need of filtering intra-domain links.
关键词：Link Analysis;Attractive Force Model;Link Weight;Web Information Retrieval
