An iterative method for personalized results adaptation in cross-language search

On today's Web, people often desire to not only retrieve results which are of relevance to their query, but for those results to be of particular relevance to them as an individual. In most personalized search systems, the scores obtained from different rankers are linearly combined to provide the personalized ranked list. Moreover, when compared to the personalization research in monolingual web search, relatively few studies extend to the cross-language domain. In this paper we investigate the personalized results adaptation problem in the context of cross-language web search. The main contribution of this research is a novel iterative ranking method based on document associations obtained from an initial ranker. The method assumes that results retrieved by non-personalized rankers and personalized rankers mutually reinforce each other, rather than being used in linear combination. The method is applied in a personalized cross-language search scenario on a semi-automatically constructed test collection and a real-world dataset. The experimental results suggest that the proposed personalized result adaptation method can produce better results than previous approaches for cross-language web search. The results also prove that the semi-automatically constructed test collection can be used as an alternative dataset for evaluation in the absence of available real-world datasets.