Strategic partnership: IBM
I was brought on quite late to this project as the lead product advisor. IBM Watson wanted to collaborate with TED to bring big data insights to its audience. The project had two outcomes:
1. to present to TED2015 attendees their own interactive profile mapped against aggregated data that compares them to other attendees sentiment in real-time; The interactive panel was able to map out talk popularity in real time right at the end of each session.
2. to extend this learning into a separate web portal that utilizes Watson's machine learning capabilities to match your sentiment with a talk that you might like - a la smart recommendation engine.
I participated mostly in the second part of the project, which unfortunately did not launch. Machine learning algorithm in 2015 was unable to correctly map talk sentiments. We learned that the spoken words, stories, etc, are given with narrative arc that's different from the written words. Watson weighted most talk sentiments as negative because speakers generally takes a substantial amount of time talking about their problem scope. It was a very educational and collaborative effort. IBM team walked away with lots of data to further train the machine. We felt pretty good about the TED2015 exhibit.
Photos are from the event.