Watson Concept Insights: A Conceptual Association Framework

Michele M. Franceschini, Livio Baldini Soares, Luis A. Lastras Montaño

Abstract: Watson Concept Insights (WCI) is a service that was recently made publicly available by IBM. WCI provides an information retrieval framework that is designed to facilitate search and exploration of text documents, and is particularly effective on sparse data sets. Its methodology consists of first defining a dictionary of concepts which are interconnected in a concept graph and then modeling a document by predicting its relevance to any given concept in the concept graph using the concepts that are directly mentioned in the document itself. This technique in effect increases the document recall for any given query, even for very sparse data sets, exposing the user to a variety of connections between their query and a data set of interest.