Overview of the TREC 2010 Entity Track

Abstract : The issue of combining (noisy) textual material (the Web) with semi-structured data (like Wikipedia or slightly more structured data sources like IMDB) is however an interesting line of research. As many data sources, and in particular those being constructed as so-called Linked Open Data (LOD), are naturally organized around entities, it would be reasonable to examine this problem in the context of entity retrieval. To foster research in this direction, we introduced the new Entity List Completion (ELC) pilot task. ELC is motivated by the same user scenario as REF, but with the main difference that entities are represented by their URIs in a Semantic Web crawl (the Billion Triple Collection). In addition, a small number of example entities (defined by their URIs) are made available as part of the topic definition. Our goal is to turn this pilot task to an "official" task in 2011.