A Non-intrusive Movie Recommendation System

Abstract

Abstract. Several recommendation systems have been developed to support the user in choosing an interesting movie from multimedia repositories. The widely utilized collaborative-filtering systems focus on the analysis of user profiles or user ratings of the items. However, these systems decrease their performance at the start-up phase and due to privacy issues, when a user hides most of his personal data. On the other hand, content-based recommendation systems compare movie features to suggest similar multimedia contents; these systems are based on less invasive observations, however they find some difficulties to supply tailored suggestions. In this paper, we propose a plot-based recommendation system, which is based upon an evaluation of similarity among the plot of a video that was watched by the user and a large amount of plots that is stored in a movie database. Since it is independent from the number of user ratings, it is able to propose famous and beloved movies as well as old or unheard movies/programs that are still strongly related to the content of the video the user has watched. We experimented different methodologies to compare natural language descriptions of movies (plots) and evaluated the Latent Semantic Analysis (LSA) to be the superior one in supporting the selection of similar plots. In order to increase the efficiency of LSA, different models have been experimented and in the end, a recommendation system that is able to compare about two hundred thousands movie plots in less than a minute has been developed.

Comparing Topic Models for a Movie Recommendation System

“…Keywords with a document frequency equal to 1 are discarded. Since, our previous work (Farinella et al, 2012) has compared tf-idf and log weighting techniques revealing that the results are very similar, in this paper we employ only the tf-idf technique for computing the weights.…”

“…The principal aim of a local repository of movies is to supply an extensive and reliable representation of multimedia that can be queried in a reasonable time. The local database of our system has been defined, as in our previous work (Farinella et al, 2012), by importing data from external repositories. In particular, we selected the Internet Movie Database (IMDb) 7 , DBpedia 8 and the Open Movie Database (TMDb) 9 .…”

“…This dimensionality in both cases has to be decided a priori. By adopting LSA, we were able to represent each plot of the IMDb database with 500 topics, instead of 220,000 keywords (Farinella et al, 2012). For LDA (which has been demonstrated working well for a number of topics over 50 (Blei et al, 2003)), after a few experimental evaluations, we decided to use 50 topics.…”

“…Starting from our previous work (Farinella et al, 2012) that highlighted the power of using the Latent Semantic Analysis (LSA) in contrast with weighting techniques (such as log and tf-idf) in suggesting similar movies, in this paper we intend to integrate in the system the Latent Dirichlet Allocation (LDA) (Blei et al, 2003) and then, by evaluating on real users we compare the performance of our system based on LSA and LDA.…”

Situation assessment for non-intrusive recommendation

“…"the possibility to disturb or to upset the user which leads to a bad answer of the user". Several works addressed this aspect as a user modelling issue and considered that a non-intrusive recommendation approach is an approach that can implicitly figure out the users' preferences and related information [45], [46], [47]. In the following sections, we present the two types of approaches that addressed intrusiveness within RSs.…”