Deb Roy, Co-Founder & CEO, Bluefin Labs
“Connecting TV to the social web”
Record everything, dev data machines/algorithms to find patters, dev visualisation so humans can see the connections and then dev them to make the world better
Was looking at how his son learnt, wired up his house for camera and microphones. Recorded everything in basement. Collected film over 3 years, (have put in strong data protections) Looked at how move through space/time in the various activities. tracked where son was in the house, got the words he was hearing, and tracked the words the son started to use. By time was 2, he had been learnt about 500 words, and they knew how they had been developed. They looked at how often a word was heard and when the son used it. So when a word used a lot, son spoke it earlier. but how about social context? The study allowed visual context to be assessed. Reveals social activity structures through time, tracking activity and then associates words heard with the activity being done. Tracking the hearing over time, gives you a landscape in space where it was heard. The new hypothesis is that the uniqueness of the wordscape means word will be learnt earlier – ie that when things happen in specific places, then the word is learnt earlier.
So context has a strong correlation with language acquisition. Took that model and started to look at TV, advertising etc. Looking at how people are commenting on TV, through social networks. You see huge growth in people talking about TV. If you look at the social expressions, in reaction to media impressions..now people create online and it can be studied. They can follow connections and where information travels They have the social web, the TV web of content, and the links between what is being said and what it being watched. This is the TV genome, the most cohesive data set of US tv and its audience. Of Google was developed to assess inbound links to content…but if you take every comment as a response to TV, then TV genome is like pagerank for TV content. They measure how many people talk about each show, Can assess the raw data, it’s not just the number that watched.
Is using this for developing insights. Looking at how to produce visualisations to share their data. Provide conversation patterns, ways to show engagement – which is not numbers, but about what is being talked about it. Does semantic analysis. Looks at audience, how they are linked etc, what types of shows people talk about.
Privacy. Only uses public data, looks at audience patterns and analyses on an aggregate level
Personal note: I worked with a media client 2 years ago to develop something very similar for them on Twitter. They wanted to track information around their shows, to be able to understand what was being said, the sentiment etc and visualise it in a way that the average show producer could get a handle on the social web.
This work is licensed under a Creative Commons Attribution-NonCommercial 4.0 International License.