Developments In Sentiment Analysis
Syntonetic CEO Peter Berg Steffensen shares his insights about the development of Moodagent, and some brief thoughts on where sentiment analysis may be heading.
One of the developments we’re currently tracking is the manifestation of more tools for understanding mood and sentiment analysis. A number of services have been popping up around this idea – Littlecosm and Tweetfeel just to name a couple – of which the most notable are trying to passively gather information about mood, aggregate this sentiment in some way, and potentially provide a layer of analysis that could result in an interesting recommendation engine. We were notified by the Winamp team that they have begun to incorporate a similar type of recommendation system for their music platform now powered by Syntonetic’s Moodagent, so we took the opportunity to speak with Syntonetic CEO Peter Berg Steffensen about these ideas. Below Peter shares his insight on the thinking behind the development of Moodagent, and some brief thoughts on where sentiment analysis may be heading.
How do you see sentiment analysis + social recommendation changing over the next 3-5 years?
As the availability of music on device, in home entertainment and in-cloud nears completion, personalization will be key to entertainment fullfilment. Capturing the personal sentiment play-by-play will allow Moodagent to build engaging experiences by bridging in relevant media assets in novel ways (how about an Emotional Weather Forecast, as Tom Waits suggests?). This applies equally to music videos and other digital assets with an audio as a natural component. Expanding this to include group sentiments in social networks is a natural extension that can serve to “guide” or inspire, especially in the physical space … I´d expect the hardware and sensor manufacturers to match this with e.g. NFC heart/mood-rate detection.
What new ideas are emerging around sentiment tracking?
Together with the University of Glasgow and mobile hardware manufacturers, Moodagent is currently working on a couple of prototypes for sentiment tracking and control for both individuals and groups, but you´ll have to wait for the details until we get them out of the lab.
We currently have systems that center around things like pushing a ‘like’ button, rating systems, or tags to catalog mood. How can these things be more passive experiences?
Moodagent can map the individual history of use, and with that group use-history, we can put the experience on remote control for one and all (applying sensors where available, of course) … but we´d like to think that we can build such engaging products that our users will want to play along.