Facebook Sentiment Engine – its not all Drama about nothing!

  2 m, 30 s

Now I’m big fan of the Drama 2.0 blog as I think he generally calls it right about most of the Web 2.0 and Social Media news around, but this time I think he’s got it dead wrong. As I’m sure a lot of you are aware the World Economic Forum took place in Davos and Robert Scoble scored a chat with Facebook Honcho Mark Zuckerberg. Amongst the usual fluff this interesting titbit came out

Facebook is, he told me, studying “sentiment” behavior. It hasn’t yet used that research in its public service yet, but is looking to figure out if people are having a good day or bad day. He said that already his teams are able to sense when nasty news, like stock prices are headed down, is underway. He also told me that the sentiment engine notices a lot of “going out” kinds of messages on Friday afternoon and then notices a lot of “hungover” messages on Saturday morning. He’s not sure where that research will lead. We talked about how sentiment analysis might lead to a new kind of news display in Facebook. Knowing whether a story is positive or negative would let Facebook pick a good selection of both kinds of news, or maybe even let you choose whether you want to see only “happy” news.

Now Drama sees this news as a excuse to poke fun at Zuckerberg but I actually see this as a really big thing and it could just be the final piece of the puzzle that makes automated sentiment analysis an accepted technology in the same way that search has. Of course at the moment Sentiment Analysis is used in several ‘Voice of the Customer’ type applications and also in various trading systems around the world but most of that analysis occurs on news data, which is generally well structured and well written but it hasn’t quite made the breakthrough to being a standard technology. As Jeff alluded to in a previous entry, there is still a lot of smoke and mirrors in the industry and whilst corporate customers get the advantages automated sentiment analysis can often bring to their business, they still have a mental hurdle to get over as to its effectiveness and ROI - of course 10 or 12 years ago Search was in much the same place! If Facebook can get it right and actually show on the site that determining sentiment enables them to do x and y better, then it makes it a win for all of us. The challenge they face is that its data tends to be a small number of words and contains both text speak and spelling errors, which tend to make a mess of automated tagging systems etc. Of course it would be even better if they opened up there (the users?) content and let the rest of us play with it as well, but I realise that’s just a pipedream - ah well maybe one day.

Categories: Social Media, Technology