Propostas de Estágio 2014/2015 - Plurianual

DEI - FCTUC
Gerado a 2024-03-29 05:16:52 (Europe/Lisbon).
Voltar

Titulo Estágio

Content-based, Movies/TV shows Recommender System Centered on Emotion

Área Tecnológica

Inteligência Artificial

Local do Estágio

CISUC

Enquadramento

Broadly speaking, recommender systems are based on at least one of the following two strategies: content-based and collaborative-based. The content filtering approach creates a profile for each user or product to characterize its nature. For example, a movie profile could include attributes regarding its genre, the participating actors, its box office popularity, and so forth. User profiles might include demographic information or answers provided on a suitable questionnaire. The profiles allow programs to associate users with matching products. Content-based strategies require gathering external information that might not be available or easy to collect.
An alternative to content filtering relies only on past user behavior - for example, previous transactions or product ratings - without requiring the creation of explicit profiles. This approach is known as collaborative filtering, which analyzes relationships between users and interdependencies among products to identify new user-item associations.
Modern consumers are inundated with choices. Electronic retailers and content providers offer a huge selection of products, with unprecedented opportunities to meet a variety of special needs and tastes. Matching consumers with the most appropriate products is key to enhancing user satisfaction and loyalty. Therefore, more retailers have become interested in recommender systems, which analyze patterns of user interest in products to provide personalized recommendations that suit a user’s taste. Because good personalized recommendations can add another dimension to the user experience, e-commerce leaders like Amazon.com and Netflix have made recommender systems a salient part of their websites.
Such systems are particularly useful for entertainment products such as movies and TV shows. Many consumers view the same movie/TV show, and each consumer is likely to view numerous different movies/TV shows. Consumers have proven willing to indicate their level of satisfaction with particular movies, so a huge volume of data is available about which movies appeal to which consumer. Companies can analyze this data to recommend movies to particular consumers.
There is however a another kind of data that could be taken into account in this analysis: emotion. In fact, emotion is a more elaborated kind of data that is beyond the “Likes” and “Don’t likes” in that it provides a considerable amount of information about the individuals and their interactions as well as about the interactions of the individuals with the environment that is not captured by the “Likes” and “Don’t likes”. This is also true when movies and TV shows are part of that environment.
Despite all the existing different kinds of data that can be used for user modeling, emotions and related mental states, such as beliefs and goals/desires, are thus a valuable aspect to take into account in order to build better user models and explore them more effectively. Building affective models of users is thus of high importance for many applications such as recommender systems. The question is how to build affective models for such consumers? To acomplish this task, one needs on the one hand a source of raw data, and on the other hand a methodology to infer emotion from that raw data.
Regarding the former problem, in this kind of environments, the usual main sources of data that can be used to infer emotional states are also valuable, but among them perhaps written verbal communication plays the most prominent role. Moreover, it is fairly easy to have access to such raw data namely not only from movies databases (subtitles, synopse, critics, etc.), but also from social media (Twitter, Facebook, etc.).
Having solved the problem of the source of data, it remains the problem of inferring emotion from that data. In other words, the question is how to identify emotions in those texts provided by social media, movie subtitles, synopse and critics? A number of approaches has been proposed lying in the area of sentiment analysis. Most of the approaches consider only the polarity of sentences or texts. However, more elaborated systems analyze texts into several emotions (usually the six basic emotions: anger, disgust, fear, joy, sadness and surprise) either at sentence or word level.

Objetivo

The goal of this internship is to develop a content-based recommender system relying on emotional data for movies/TV shows. It comprises:
(i) building affective user models of the consumers of movies/TV shows
(ii) classifying movies/TV shows according to their emotional features;
(iii) reccommend movies/TV shows based on the matching between affective consumer models and movies/TV shows emotional classification.

Plano de Trabalhos - Semestre 1

1- State of the Art [Sept – Nov]
2- Analysis and Specification [Dec]
2.1- Definition of System Requirements
2.2- Use Case Definition
2.3- Design and Specification
3- Prototype Development [Dec – Feb]
4- Thesis Proposal Writing [Dec – Feb]

Plano de Trabalhos - Semestre 2

5- Prototype Improvement [Mar – May]
6- Experimental Tests [Jun – Jul]
6.1- Functional Tests
6.2- Performance Tests
7- Thesis Writing [Jun – Jul]

Condições

The work will be developed in one of the labs of the Cognitive and Media Systems.

Observações

Basic knowledge of programming and artificial intelligence are required.

Orientador

Luís Macedo
macedo@dei.uc.pt 📩