Sentiment analyis: one to watch

10 December 2005

For any company needing to listen to the blogosphere, sentiment is key. It’s one thing to know that there are conversations taking place about your sector or company; it’s quite another to realise that 74% of them are negative (or positive) towards you.

I’ve always been sceptical about monitoring services that claim to have technology that can automatically identify sentiment, particularly because you have to pay for them. Which is why I was interested to hear about Opinmind (The Basement via prdigest).

This new “search engine with a twist” uses an “opinion analytics engine” (called Opal™) to find what people like and don’t like. It then displays results in two columns – negative and positive – along with a Sentimeter™ which shows the relative number of both kinds of mention.

There’s another undocumented feature highlighted in their blog. To compare two terms, you can just insert “vs” between them, like this:

PR vs Advertising – and judging by the result, it must be accurate ;-)

I’m sure their engine doesn’t get it right all the time (as they admit, “the complexities of written language and of blogger grammar (or lack thereof) occasionally throw a wrench into our opinion analytics engine”, but for a free service this is pretty good. As long as it trips up on judging positive and negative in equal proportions, then as a dip test it should be pretty reliable.

Compared to other search engines, their database is pretty small (seems to be mainly indexing LiveJournal, Xanga and MSN Spaces at the moment), so I encourage you to submit your blog or xml feed at the bottom of the results page.

Now how about an API, guys?

Share and Enjoy:
  • Digg
  • Facebook
  • Google Bookmarks
  • email
  • NewsVine
  • Slashdot
  • StumbleUpon
  • Twitter

3 Responses to “Sentiment analyis: one to watch”

  1. Euan


  2. lexicon

    How about "merry christmas" vs. "happy holidays"?

  3. Niall Cook

    Happy holidays is the clear winner at the moment.

Leave a Reply