Abstract

This paper describes the DAI approach to the TAC-KBP 2011 Entity Linking Task, which this year introduced the subtask of creating novel knowledge base (KB) entries from name mentions referring to entities unknown to the KB (NIL clustering). For the entity linking task, our system uses disambiguation features that are based on KB information as well as on semantic similarity relations between named entities identified in the document context and in KB entries. To solve the NIL clustering task we implemented a three-stage approach which aims to improve upon an initial name mention string similarity clustering by introducing separate clustering steps for ambiguity and synonymy resolution. We describe implementation details of our system, and present a preliminary analysis of the results.

@inproceedings{ploch11b,
author = {Danuta Ploch and Leonhard Hennig and Ernesto William De Luca and Sahin Albayrak},
title = {DAI Approaches to the TAC-KBP 2011 Entity Linking Task},
booktitle = {Proceedings of the Text Analysis Conference},
year = {2011},
}