Syncretic matching
Sumant Kulkarni, Srinath Srinivasa, Tahir Dar · 2018
In several document matching applications like comparing across judgments, patent claims or movie plots, conventional bag-of-words models are insufficient. Bag of words are useful for computing lexical similarity; while in this case, there is a need to understand similarity with respect to the underlying narrative or "story." We call this the Syncretic matching problem. While bag-of-words can be enhanced by using techniques like dimensionality reduction or topic models, the syncretic matching problem is more involved. It requires modeling the underlying semantic "story" and comparing structural similarities across stories. In this paper, we address the problem of narrative similarity computation for given pair of input documents. The approach utilizes a general knowledge base in the form of a term co-occurrence graph (TCG) computed from all articles in Wikipedia, to help in creating a story model for comparison.