I came across a project at the Innovation labs at the recent IBMLotusphere conference whereMarie Wallace, a researcher and Analytics Strategist from IBM in Dublin, Ireland, was discussing the issues of measuring analytics in online social environments. In particular, she described her research experiment to apply the ideas of semantic analysis to what people discuss in a social site like Twitter. I consider this area as a fundamental new business area in the technology field with many particularly challenging technical aspects, on the scale of finding the next Google-sized opportunity. To understand the significance, we need to skim the surface of the field of social analytics.
To begin, a good many influence determination tools (e.g.,Klout,LinkedIn’s InMaps social graph) look at influence in terms of interaction patterns between people and their individual social networks, or networks of friends and followers. These tools look at the dynamics of the interaction primarily rather than content. For example, it could determine how far an individual posting that you make will spread across the people that follow your comments on Twitter, or even across millions of users in Twitter itself. Similarly it could count how often on average something you say is shared and propagated by others. They can even rank you relative to the rank of other people that you in turn follow—this is recursive, they themselves are ranked relative to yet others they follow and interact with.
These types of metrics can certainly tell us something about behavior of people who listen to you. If you think of this information in terms of audience, now every single person can have their own personal Nielsen ratings—the metric made famous by the company focused on the Television industry, and used heavily as a guide to advertising dollar worth of content.
However, these metrics primarily focus on the patterns of the interaction, rather than the content of what is being discussed. You can do a great of analysis based on the interaction pattern alone, but the next step beyond is to look at the content itself–the realm of semantic analysis. Rather than theinteraction pattern analysis which is essentially a counting or quantitative problem, we are now looking for the deeper meaning, a qualitative problem.