The authors investigate two publicly available Web knowledge bases, Wikipedia and Yago, in an attempt to leverage semantic information and increase the performance level of a state-of-the-art coreference resolution engine. They extract semantic compatibility and aliasing information from Wikipedia and Yago, and incorporate it into a coreference resolution system. The authors show that using such knowledge with no disambiguation and filtering does not bring any improvement over the baseline, mirroring the previous findings (Ponzetto & Poesio, 2009). They propose, therefore, a number of solutions to reduce the amount of noise coming from Web resources: using disambiguation tools for Wikipedia, pruning Yago to eliminate the most generic categories and imposing additional constraints on affected mentions. The evaluation experiments on the ACE-02 corpus show that the knowledge, extracted from Wikipedia and Yago, improves the system’s performance by 2-3 percentage points.
TopCoreference is a complex phenomenon and therefore a robust and reliable approach to the problem should address numerous linguistic and common-sense aspects of the task. Previous studies have investigated possibilities for extracting such knowledge from WordNet (Harabagiu, Bunescu, & Maiorano, 2001, Huang, G., W., & A., 2009), Wikipedia (Ponzetto & Strube, 2006) or large text corpora (Haghighi & Klein, 2009, Bean & Riloff, 2004, Garera & Yarowsky, 2006, Yang & Su, 2007).
In the present study, we investigate possibilities of integrating information extracted from Wikipedia and Yago into a coreference resolution system. It has been shown that, even though at earlier stages web knowledge bases might be a source of valuable information (Ponzetto & Strube, 2006), the expansion of such resources inevitably leads to an increase in the amount of noise, making them hardly usable for our application (Ponzetto & Poesio, 2009).
The research line on Wikipedia related to our work is the automatic annotation of terms in a plain text with links to Wikipedia pages. In fact, it is a WSD task because its goal is to link a term in a sentence to the Wikipedia concept that best expresses its sense. Some well-known approaches to this task include (Csomai & Mihalcea, 2008) and (Milne & Witten, 2008). They perform the so-called wikification of the document, that is they first identify the main concepts in a text and then annotate them with links to Wikipedia pages.