Delft University at the TREC 2009 Entity Track: Ranking Wikipedia Entities

Entity ranking is a novel TREC task introduced this year and posing challenges similar to those already well-known in web retrieval research community. We present a system which leverages a number of popular web retrieval techniques, utilizes existing knowledge bases and relies on various task-specific heuristics to produce high quality runs. Since we had no training queries with relevant web-pages contained in Category B part of ClueWeb09 collection, we focused on using various strategies rather than on using one kind of approach with different parameter settings. However, in three of four submitted runs we treated Wikipedia part of the collection as the main source of evidence about relevance of entities that can be found on the Web.