Contextualized query sampling to discover semantic resource descriptions on the web

摘要

Resource description extracted by query-sampling method can be applied to determine which database sources a certain query should be firstly sent to. In this paper, we propose a contextualized query-sampling method to extract the resources which are most relevant to up-to-date context. Practically, the proposed approach is adopted to personal crawler systems (the so-called focused crawlers), which can support the corresponding user’s web navigation tasks in real-time. By taking into account the user context (e.g., intentions or interests), the crawler can build the queries to evaluate candidate information sources. As a result, we can discover semantic associations (i) between user context and the sources, and (ii) between all pairs of the sources. These associations are applied to rank the sources, and transform the queries for the other sources. For evaluating the performance of contextualized query sampling on 53 information sources, we compared the ranking lists recommended by the proposed method with user feedbacks (i.e., ideal ranks), and also computed the precision of discovered subsumptions as semantic associations between the sources.