TVPulse: Improvements on detecting TV highlights in Social Networks using metadata and semantic similarity

Afonso Vilaça, Mário Antunes, Diogo Gomes, "TVPulse: Improvements on detecting TV highlights in Social Networks using metadata and semantic similarity", Proc. 14ª Conferência sobre Redes de Computadores, Évora, Nov 2015

Tags: Big Data, social networks, text mining

Abstract

Sharing live experiences in social networks is a growing trend. That includes posting comments and sentiments about TV programs. Automatic detection of messages with contents related to TV opens new opportunities for the industry of entertainment information.
This paper describes a system that detects TV highlights in one of the most important social networks - Twitter. Combining Twit- ter’s messages and information from an Electronic Programming Guide (EPG) enriched with external metadata we built a model that matches tweets with TV programs with an accuracy over 80%. Our model required the construction of semantic profiles for the Portuguese language. These semantic profiles are used to identify the most representative tweets as highlights of a TV program. Measuring semantic similarity with those tweets it is possible to gather other messages within the same context. This strategy improves the recall of the detection. In addition we developed a method to automatically gather other related web resources, namely Youtube videos.

Information

Conference: 14ª Conferência sobre Redes de Computadores in Évora