Improving web search results using affinity graph

Abstract

In this paper, we propose a novel ranking scheme named Affinity Ranking (AR) to re-rank search results by optimizing two metrics:(1) diversity --which indicates the variance of topics in a group of documents; (2) information richness --which measures the coverage of a single document to its topic. Both of the two metrics are calculated from a directed link graph named Affinity Graph (AG). AG models the structure of a group of documents based on the asymmetric content similarities between each pair of documents. Experimental results in Yahoo! Directory, ODP Data, and Newsgroup data demonstrate that our proposed ranking algorithm significantly improves the search performance. Specifically, the algorithm achieves 31% improvement in diversity and 12% improvement in information richness relatively within the top 10 search results.

Mining and Explaining Relationships in Wikipedia

“…Consequently, disjoint paths produce higher NDCGs than dependent paths, in most cases. The experimental results are similar to that removing redundancies in search results improves recall and precision in information retrieval system [5]- [7].…”

“…Therefore, recent document information retrieval methods [5]- [7] adopt an idea that redundant information should be minimized in the top-ranked documents by removing documents similar to a higher ranked documents. For example, given a query "foreign relationships of the USA," a set of the top-ranked documents should cover relationships between the USA and various countries, the set should not contain a number of similar documents explaining the relationship between the USA and a certain country.…”

“…In Sect. 5 most of the top-20 web pages searched by a keyword search engine using a query "s t" are inadequate for understanding the relationship between s and t.…”

“…An idea of displaying disjoint paths is that the same or similar paths should not appear multiple times in the paths shown for explaining a relationship. A similar idea is widely accepted in the field of document retrieval: a search result should not contain same or similar documents [5]- [7]. However, to the best of our knowledge, there has been no study that mines disjoint paths for explaining a relationship on an information network.…”

Clustering Web Search Results with Maximum Spanning Trees

“…Similarity functions have been used to measure the diversity among documents and between document and query [5]. Other techniques include the use of conditional probabilities to determine which document is most different from higher-ranking ones [9] or use affinity ranking [46], based on topic variance and coverage. More recently, an algorithm called Essential Pages [38] has been proposed to reduce information redundancy and return Web pages that maximize coverage with respect to the input query.…”