Using fuzzy logic to estimate user interests in multiscreen viewing situation

This paper describes a method to extract user interest keywords for user profiling applicable to various recommendation systems from browsed web pages during multiscreen viewing, i.e. watching TV while browsing web pages related to the TV program on smartphones or Tablet PCs. Our method extracts keywords from browsed web pages and computes interest weights, which is based on the user´s natural behaviors of “browsed”, “bookmarked; like social bookmarking” or “no reaction; user did not browse it although it was shown”. Ranking the keywords in order of interest weights allows us to estimate the interest keywords. Experiments show that our method has better potential in filtering out non-useful keywords and thus is superior to the basic TFIDF approach.