Media content personalisation brokerage platform

Resum

Media content personalisation poses a significant challenge to media contents producers, distributors and viewers. The goal of media content producers is to place their media items at the highest price possible and, the goal of a content distributor is to strengthen the relationship with existing customers and increase the number of customers by offering the most appropriate media content from a multitude of sources, while taking the viewer profile into account. The viewers are unable to search, find and choose their preferred content in near real time due to the size of the search space both in terms of the number of offers and the diversity of sources. The nature of relationships between contents producers and distributors is Business-to-Business (B2B), whereas the nature of the relationship between distributors and viewers is Business-to-Consumer (B2C). The described problem opens several issues, namely, in terms of profiling, recommendation and contents brokerage. Not only it is necessary to develop a profiling mechanism capable of building and maintaining user and item models over time, but also to find recommendation algorithms capable of processing the different types of user feedback in order to provide the best recommendations in near real time. Finally, the media contents provider and distributor businesses need to delegate the partner lookup, invitation and acceptance to negotiate and enforce the provision of the recommended objects to a brokerage service. In this thesis we tackle the problem of near real time media content personalisation. In order to achieve this, we designed a platform meeting the above mentioned requirements. Our contributions are focussed on the three main designed components: (i) Profiling Service; (ii) Recommendation Service; and (iii) a Multi-Agent System (MAS) Brokerage Platform.