Utilize este identificador para referenciar este registo: http://hdl.handle.net/10400.22/10778
Título: A Hybrid Approach for Personalized News Recommendation in a Mobility Scenario Using Long-Short User Interest
Autor: Viana, Paula
Soares, Márcio
Palavras-chave: News recommendation system
Geolocation
Long-short preferences
Hybrid recommender
Data: 2017
Editora: World Scientific Publishing
Resumo: Access to information has been made easier in different domains that range from multimedia content, books, music, news, etc. To deal with the huge amount of alternatives, recommendation systems have been often used as a solution to filter the options and provide suggestions of items that might be of interest to an user. The news domain introduces additional challenges due not only to the large amount of new items produced daily but also due to their ephemeral timelife. In this paper, a news recommendation system which combines content-based and georeferenced techniques in a mobility scenario, is proposed. Taking into account the volatility of the information, short-term and long-term user profiles are considered and implicitly built. Besides tracking users’ clicks, the system infers different levels of interest an article has by tracking and weighting each action in the system and in social networks. Impact of the different fields that make up a news is also taken into account by following the inverted pyramid model that assumes different levels of importance to each paragraph of the article. The solution was tested with a population of volunteers and results indicate that the quality of the recommendation approach is acknowledged by the users.
Peer review: yes
URI: http://hdl.handle.net/10400.22/10778
DOI: 10.1142/S0218213017600120
Aparece nas colecções:ISEP - DEE - Artigos

Ficheiros deste registo:
Ficheiro Descrição TamanhoFormato 
ART_PViana_DEE_2017.pdf1,87 MBAdobe PDFVer/Abrir    Acesso Restrito. Solicitar cópia ao autor!


FacebookTwitterDeliciousLinkedInDiggGoogle BookmarksMySpace
Formato BibTex MendeleyEndnote 

Todos os registos no repositório estão protegidos por leis de copyright, com todos os direitos reservados.