“Julia Gillard was in my dream last night. It was quite bizarre.” One point for Julia.
“Tony Abbott‘s laugh is a little bit terrifying, but at least he can laugh.” One point for Tony.
This is how computers are determining whether our political leaders are in or out of favour with the public. Chunks of text are fed into a program and each tweet, Facebook status, blog post or news article is given a positive or negative stamp. It’s all totalled up and you end up with a single figure between -100 and 100. They call it sentiment analysis.
But are computers smart enough to understand the nuances of language, from cultural differences to sarcasm and irony? Or is sentiment analysis nothing but snake oil for marketing agencies to sell to the many corporations eager to get on the social media bandwagon, but unable to even locate the wagon in the car park?
Brand management is where most sentiment analysis is currently happening. Chocolate milk makers, fearing their product will be seen as a daggy school-related drink, are monitoring kids on social networks to see how their product is being talked about. Car companies are monitoring how many Facebook groups have been setup about their brand – both positive and negative – as well as how many people are in the groups and whether they’re making positive or negative statements.
Corporations love being talked about, and being talked about in social media circles is an enticing way to be in the headspace of the lucrative Gen Y-ers. There is a rush to know how to react when things go wrong, or to control what is being said. Sentiment analysis offers a foothold into this area, a modern day media monitoring service sans humans. The truth is that the best sentiment analysis providers claim to provide only a 70-80 per cent level of accuracy.
Continues @ http://www.abc.net.au
Related articles by Zemanta
- The Problem with Automated Sentiment Analysis (futurelab.net)
- Perspective of sCRM (digitalassetmanagement.org.uk)
- Sentiment in an Unsentimental Manner (arnoldit.com)
- Using Sentiment Analysis to Measure Social Media Engagement (customerthink.com)
- Sentiment vs. message in social media – which do you value more? (prbreakfastclub.com)