Adaptive contents for interactive TV guided by machine learning based on predictive sentiment analysis of data

Abstract

This paper describes a new proposal for interactive television which is an answer to a continuous change in the traditional television as consequence of the inclusion and evolution of the digital social networks, the Internet and the different elements of the digital age. The digital evolution has encourage the interaction of the viewers with the content and also increases the need to evolved the content, the methods, formats, tools and architectures to adapt the content to the sentiment expressed by the viewer while watching a show. The present paper contains the following objectives: The first objective is to create guidelines that can be used to construct adaptive contents for television, which can be modified in real time by the production team or the director of the show. The second objective is to develop applications that allows to obtain, collect and analyze the sentiment inside of the expressions, data or opinions of the viewers, who interact with the show through social networks or communication channels as: Facebook, Twitter, Instagram and WhatsApp. The third objective is to develop a machine learning to predict the preferences of the viewers, generating options and changes in the sequence of the scenes of the TV show that will be broadcasted in real time. All the objectives explained above are applied to two TV shows which are different in the content but share the live condition. During the broadcasting of the show, the guidelines are applied, the results are obtained, analyzed and the final result is more participation of the viewers and a better perception of the content. As a result of the research and the application in real life of the proposal, this paper contributes with an alternative solution for interactive TV where a viewer can interact with the show and the production team can modify the content according to what the viewers express and expect to watch based on an analysis of sentiment of data using a machine learning.