MOVSOM-II - analyzis and visualization of movieplot clusters 2004-07-27


Download the paper as a PDF

Abstract

In the paper "Movie clustering using the self-organizing map: MOVSOM" we presented MOVSOM, a fully automatic movie recommender system with no need of user preferences whatsoever. In contrast to movie recommender systems like MovieLens, that are depended on user input in form of rankings of movies or IMDB's recommending system that relies heavily on users recommendations and keywords in order to be able to recommend movies, MOVSOM doesn't require any input. Our aim was to cluster movieplots and visualize the results in a user friendly way and make it possible for a user to select a movie they have seen and then be recommended movies that are similar to that movie in some way. Our belief that this was possible was due to WEBSOM, which is a system for clustering documents based on their content. In this paper we present MOVSOM-II, a successor to MOVSOM, but with better visualisation, preprocessing techniques and analyzing of the result.

Bibtex

@unpublished{gabrielsson04movsom,
  author={Stefan Gabrielsson and Sam Gabrielsson},
  title={{MOVSOM-II} - {A}nalyzis and {V}isualization of {M}ovieplot {C}lusters},
  note={Uppsala University, Department of Information Technology},
  year={2004}
}