Mining future spatiotemporal events and their sentiment from online news articles for location-aware recommendation system

摘要: The future-related information mining task for online web resources such as news articles and blogs has been getting more attention due to its potential usefulness in supporting individual's decision making a world where massive new data are generated daily. Instead of building data-driven model predict the future, one extracts future events from these with high probability that they occur at time specific geographic location. Such spatiotemporal can be utilized by recommender system on location-aware device provide localized event suggestions.In this paper, we describe systematic approach web; particular, articles. In our application context, valid is defined both spatially temporally. procedure consists two main steps: recognition matching. For step, identify resolve toponyms (geographic location) temporal patterns. matching perform disambiguation, de-duplication, pairing. To useful guidance, attach each sentiment linguistic variable: positive, negative, or neutral, so may use extracted recommendation purposes form "avoid Event A" location L T" "attend B" based sentiment. identified location, pattern, variable, title, key phrase, article URL. Experimental results 3652 21 sources collected over 2-week period Greater Washington area used illustrate some critical steps procedure.