ualberta at TAC-KBP 2012: English and Cross-Lingual Entity Linking

On one hand, the proliferation of the Web has generated massive information in an unorganized way and is still growing in an accelerating pace. On the other hand, structured and queryable knowledge bases are very difficult to construct and update. Automatic knowledge base construction techniques are greatly needed to convert the rich Web information into useful knowledge bases. Besides information extraction, ambiguities about entities and facts also need to be resolved. Entity Linking, which links an extracted named entity to an entity in a knowledge base, is to solve this ambiguity before populating knowledge. In this paper, we describe ualberta’s system for the 2012 TAC-KBP English and Cross-Lingual Entity Linking (EL) task, and report the result on the evaluation datasets.