Recommending paragraphs of wikipedia pages as a travel guide

Abstract

This paper proposes a method to recommend paragraphs of Wikipedia pages as a travel guide. This method helps tourists (users) read attractive descriptions of Wikipedia pages during their travel. In order to rate paragraphs, importance, non-redundancy, and novelty of the paragraphs are evaluated based on a Tf-Idf manner. Especially, novelty is done by user's experience estimated from geo-tagged tweets located to places that the user visited in the past. As the results of the experiments, we confirmed the geo-tagged tweets reflected user's experience and the recommendations exceeded the expectation of MAP criteria.