Mining Coreference Relations between Formulas and Text using Wikipedia

In this paper, we address the problem of discovering coreference relations between formulas and the surrounding text. The task is different from traditional coreference resolution because of the unique structure of the formulas. In this paper, we present an approach, which we call ‘CDF (Concept Description Formula)’, for mining coreference relations between formulas and the concepts that refer to them. Using Wikipedia articles as a target corpus, our approach is based on surface level text matching between formulas and text, as well as patterns that represent relationships between them. The results showed the potential of our approach for formulas and text coreference mining.