Unsupervised Context-Driven Recommendations Based On User Reviews
Francisco J. Ribadas · 2017
In this work we present Rich-Context, a context-driven recommender system that extracts contextual information using topic modeling without the need to define keywords. Our system uses the mined context to produce recommendations. We propose a methodology to measure the quality of context topic models along with a novel way to represent context that allows it to be used as side-information in a recommendation engine. Results show that Rich-Context makes more accurate predictions than five well-established recommendation algorithms.