Big Data for the TV Industry: Tune into the Backchannel

This talk is about the tweet back channel. According to the presenters, 86% of people that have a tablet or smart phone use it to comment on TV shows on social media. This leads to the fact that up to 40% of tweets in the time line can be about TV shows at prime time. The guys at Mesagraph have built a platform to analyze tweets on TV shows.

The platform works like this:

Entity extraction is important, because people can use different ways to refer to the same thing, like Sarko for Sarkozy. Or the other way around: Hollande can be the name of a country or the name of the new president of France.

Challenges that the TV industry faces today are the changing way in which people watch TV and it is hard for the industry to understand and anticipate to this change. Based on tweets, the guys of MesaGraph could analyze what people watch the same type of shows. As you can see in this visualization, there are clear clusters in which the tweets are grouped, meaning that people tend to stick to the same type of show.

A TV show is an event by itself, however, also within a TV show, there are different events. MesaGraph tries to detect those in-show events by looking at tweets. Again they perform entity detection and if entities occur together, it means that something is happening with both entities. If there are a lot of this type of tweets, this might mean that an event is going on.

Finally, MesaGraph can also monitor celebrities on Twitter, to prevent real-life troubles similar to what happened to Rihanna recently, when she attracted a lot of fans to a Paris train station by tweeting she was going to there.