You've probably heard of sentiment analysis, a technique that allows computers to determine whether a comment is positive or negative about a subject. You can probably imagine a few uses for this kind of technology, such as culling social media for mentions of your brand, but you might be surprised about how much effort Google is putting into this area, including recently purchasing a startup that specializes in sentiment analysis.

Google has already made public its use of sentiment analysis for organic search. You should see its use of this technique only increase.

If you think about the way search ranking algorithms work, there are only two areas that matter: what happens on the page and what happens off the page. The on-page factors— how many instances of the search keyword appear on the page or whether they are in the title or just the body of the page—are easily gamed by spammers. They are still important, but for competitive keywords, they've never been enough. You've always had to win the off-page game, which has traditionally been about the number and quality of links to your page.

Image representing Google as depicted in Crunc...

Image via CrunchBase

There are a few problems with this approach, as successful as it has been for Google. One is that even links can be gamed, through link farms, comment spam, paid links, and many other techniques. Google has spent years battling that problem and is increasingly using other factors in addition to links to determine the quality of the page, including social media activity.

But one problem with all of these popularity techniques is that people sometimes talk about you because they hate you. And it doesn't make sense that Google ought to rank your page higher if you are hated, at least not in all circumstances. For that reason, Google is working more and more with sentiment analysis, so that they can know a bit more about the tone of each link and each social media mention.

Simple sentiment analysis uses the lexicon—the words used—to determine which statements are positive and negative, but it isn't all that accurate. Much of the time, we speak in sarcasm ("That movie was a great use of my time") or we don't use any telltale words in our opinion ("This is the Edsel of cell phones"), so smarter techniques use machine learning to match up a small set of human-coded correct answers to determine patterns (perhaps that the word "Edsel" is almost always part of a negative statement).

Google will undoubtedly have many uses for highly accurate sentiment analysis (Google Alerts that highlight negative comments?) but they continue to try to figure out what makes a high quality page, so sentiment analysis will be a growing part of search.

Enhanced by Zemanta

February 3, 2011





Mike is an expert in search marketing, search technology, social media, publishing, Web personalization, and Web metrics, who regularly makes speaking appearances.

Mike's previous appearances include ClickZ Live, RKG Summit, Ticket Summit, Webdagene, the CiTE conference, and the Forrester Marketing Conference.

Mike also founded and writes for the Biznology newsletter and blog, is the co-author of the best-selling Search Engine Marketing, Inc., and sole author of Do It Wrong Quickly, named by the Miami Herald as one of the 11 best business books of 2007.






Comments(4)

Fascinating. This is where managing a company's online presence is going to be even more essential going forward in Internet Marketing.

With the Internet, we have the awesome ability to find out what people are saying about our companies. Unfortunately many companies don't take advantage of this sometimes harsh but great information. Sure, it's tough to read consumer complaints on a forum, but imagine what you could do with that information! Shouldn't we be taking those criticisms and addressing them, sometimes even personally? Forget surveying people, many are happy to share their experience with everyone in their circle now via the Internet, sometimes instantaneously with the increased use of mobile.

Good PR isn't just in the newspapers anymore. It's everywhere - social media, forums, blog posts, and more. And it's propagated by everyone - not just journalists. The same goes for bad PR. As search engines become more adept at reading into nuances in writing, managing a company's online presence won't just be a way to better improve, it'll be a means of survival.

Thanks for the great post. It spawned a lot of ideas in my head not only about companies managing their online presence but also individuals managing their online presence.


Leslie

Thanks for the great post!

I think it is increasingly important to get good knowledge out on the subject of sentiment analysis, so that corporate clients would be able to understand their social media analysis data better.

Also, it is quite interesting to see where exactly Google will end up using these forthcoming results. Map-based sentiment analysis to show where really good and really bad places (e.g. restaurants) are nearby would be great - especially if it worked within the Google Maps ecosystem of users.

Sentiment Analysis is key to companies who worry about their brand, and should be important to new companies as well.

Many times one doesn't remember specific details (even web addresses) but will remember a tweet or a comment about your company or brand.

Keep it positive and deal with negative sentiment professionally.

Well, glad to see that Google put an end to "All press is good press" idea. I wonder why they didn't figure this out earlier. While the related post at Google blog makes for a good narrative, I doubt the first time they heard this was during that interview the author referred to.
But I do get a bit of an upgrade suggestion for Google...
They do seem to be always on the offensive, meaning, to punish evil. I wonder how it would be if they also do an algo update that will search for love comments and reward sites that have a lot of positive feedback, both on sites and on Twitter/Facebook/Etc. Then companies will have a tangible incentive to overdeliver and outperform in good customer service.
Anyway, thanks for the great article and the added value of providing the link to google's blog.

Comments closed after 30 days to combat spam.


Search Engine Guide > Mike Moran > How is Google feeling about sentiment analysis?