Microsoft Social Listening has grown to more than a thousand customers so far, and the product will be making a leap in its sentiment analysis coverage in 2015. The product will expand from six languages (with Italian to be added in the next month or so) up to nineteen as the product shifts to using more machine learning technology and less natural language processing.
A lot has changed for Microsoft Social Listening in the year and half since Microsoft acquired Netbreeze. The team, which still includes founder Francois Ruf, has migrated the product to Azure, utilizing a range of the cloud platform’s IaaS and PaaS capabilities. And they now have more than a thousand customers on the platform and an updated API layer with three levels of connectivity.
Social Listening will continuing to grow in usage, Ruf says, and the new architecture is in good shape to support the additional computing resources that will be needed. But another costly requirement for Social Listening is the maintenance and expansion of its sentiment analysis engine. Moving to a machine learning (ML) approach to augment or substitute for the more traditional natural language processing (NLP) ought to rapidly expand the product’s reach.
“NLP requires quite a lot of investment but gives you a lot of positions,” Ruf says. Some languages will begin to use a mix of NLP and ML. Others will be measuring sentiment strictly through ML.
“On ML, it’s just about how many labelled data points you have to train the model,” says Ruf. “Quality won’t differ too much. For example, there are differences in emotions that can be detected with ML. [It] can be done to make it as good as NLP.”
NLP requires computational linguists to work in…click here to read the rest of this article