Personal Content Recommender Based on a Hierarchical User Model for the Selection of TV Programmes

Abstract

In this paper we present our approach to user modeling for a personalized selection of multimedia content tested on a corpus of TV programmes. The idea of this approach is to classify content (TV programmes) based on the calculation of similarities between the description of content and the user model for each description attribute. Calculated similarities are then combined into a classification decision using the Support Vector Machines. The basis for the calculation of similarities is a hierarchical structure of the user model, overlaid upon a taxonomy of TV programme genres. Preliminary results show that it works well with a varying quality of content descriptions including incomplete genre classification and arbitrary number of description attributes. The evaluation of the system performance was based on content described using the TV-Anytime standard, but the approach can be adapted for search of other types of content with multi-attribute descriptions.

Using affective parameters in a content-based recommender system for images

“…Each of these steps can be tackled in a number of different ways, which are summarized in Table 1. The majority of early recommender systems, especially movie recommenders, used metadata fields provided by content producers via databases like imdb.com for the description of items and users (see the work carried out by Basu et al 1998;Pogačnik et al 2005 and surveys by Adomavicius and Tuzhilin 2005;Burke 2002). Typical examples of such content producers' metadata are genre, actors, subject matter, etc.…”

“…The GM set used in our comparative study is composed of the genre g and the average watching timet w of the item h. Both attributes are widely used in recommender systems (Adomavicius and Tuzhilin 2005;Pogačnik et al 2005;Kim et al 2005) and thus suitable for our comparison. The genre was set manually and was chosen from a set of ten available genres.…”

“…In contrast to the most used user modeling approach, as employed by Pogačnik et al (2005) (we used a simplified version of it for illustration purposes in Fig. 3) or surveyed by Adomavicius and Tuzhilin (2005), we propose to model the users' preferences toward emotive states with a user profile-a data structure that is the result of training the ML algorithm based on past ratings (e.g.…”

“…The set of evaluated classifiers was = {AdaBoost, C4.5, NaiveBayes, SVM}. We chose the classifiers based on the following criteria: (i) related work (Adomavicius and Tuzhilin 2005;Pogačnik et al 2005;Lew et al 2006), (ii) to cover different families of ML algorithms and (iii) based on the visual inspection of the dataset. We performed the visual inspection of the data on plots like the one shown in Fig.…”

“…We used the measures precision P, recall R and F-measure F β (where β = 1) as defined by Herlocker et al (2004) and commonly used in the evaluation of recommender systems (Herlocker et al 2004;Pogačnik et al 2005;Adomavicius and Tuzhilin 2005;Witten and Frank 2005). The precision tells us the rate of relevant items among all the selected items.…”