One small step for computational sentiment analysis, one big leap for social media measurement…

June 1, 2010

Sentiment around a brand online is held up as one of the key cornerstones of any social media measurement framework. Just driving up the volume of chatter around your brand is no good – if that chatter is all about what a terrible organisation you are.

But, one of the biggest challenges facing the industry has been how to capture the full range of that positive and negative feeling. The online world is a large and sprawling place. There are 100s of comments posted every day about a company’s products, services and public image.
There are a huge number of tools out there that attempt sentiment analysis in one form or another – Radian 6 being one of the more commonly used ones. Its creators are pretty transparent about its limitations on the sentiment front (allowing you to manually override the ‘score’ given by the tool’s algorithm).

Industry experts estimate that machine analysis is right about 60% of the time (love and hate are easy, but anything in between is not). Most of us end up conducting manual analysis, sampling multiple posts, instead. However, in New Scientist’s latest edition, there’s a shining beacon of hope for all those spending hours pouring over the 500th blog post of the day. A team from Israel’s Hebrew University of Jerusalem have created a program that achieved an 80% (give or take a few percent) success rate when it came to measuring sarcasm on Amazon and Twitter.

You can read the article in more detail here, and we’ll be following the progress of Ari Rappoport and co. with great interest over the coming months…