I did a search on “starbucks” and on “wal-mart” as part of a little experiment. Both sets of results could be characterized as sweet and sour. Sure, both brands are well represented at the top of the pile, with Wal-Mart even getting a universal-search-style Google News result (complete with image) in the middle of the page.
On the Starbucks search, what jumped out more than stock quotes, news results, and images were the Starbucks Gossip and I Hate Starbucks sites. Starbucks Gossip says it’s “Monitoring America’s favorite drug dealer,” while the “I Hate Starbucks” site is just that: a hate site.
For Wal-Mart, it was Wal-Mart Watch that jumped out the most. The description snippet states, “a nationwide campaign to reveal the harmful impact of walmart on american families.”
Those sweet and sour results compete for searchers’ attention.
SEO (define) techniques can help present a brand in the most positive light. But those same techniques can also discredit a brand. Problem is, search engines don’t know the difference between the sweet and the sour.
Search engines have no problem using technical wizardry to get results relevant to the searcher’s query. But they don’t understand the results’ meaning. Take two pages: one says “Mike loves Tatiana,” and the other says, “Tatiana loves Mike.” From a keyword point of view, the pages are virtually identical. Yet the meaning of each page is entirely different.
The computational treatment of opinion, sentiment, and subjectivity has attracted a great deal of attention in research circles. Getting the actual gist or meaning of a page isn’t a new problem. But with so much user-generated content being uploaded to the Web, reputation management becomes increasingly more important to online marketers.
At search engines, automatic text categorization provides the bulk of the heavy lifting in determining document topic. Text can then be sorted according to subject matter, such as sports vs. politics. But with so much user-generated content in blogs, forums, and review sites, the crucial element of such postings is the sentiment or overall opinion toward the subject matter.
Blogs, forums, and review sites can provide all manner of insight into a target audience. But knowing whether such comments and reviews are positive or negative is imperative.
How nice it would be for marketers to employ dozens, even hundreds, of people to scour the Web looking for references to products and services and reporting on the good, bad, and ugly. How much nicer it would be if a bot could do it for us.
Imagine a public relations tool of the future: it would crawl the Web, extracting text from documents and ranking the results in order of sentiment. But it seems there’s a long way to go.
As one sentiment analysis researcher commented, “The sentence, ‘How could anyone sit through this movie’ contains no single word that’s obviously negative.” Yet it’s an uncomplimentary remark to make about a movie.
Sentiment analysis has specifically been proposed as a key enabling technology in e-rulemaking (define), allowing an automatic analysis of people’s submitted opinions. Research has also focused on determining the political leaning of a document or author (e.g., liberal vs. conservative). But that may be a little too Big Brother-ish.
I came across an interesting experiment carried out by Cornell University researchers last year. “Get out the vote” investigates whether it’s possible to determine support or opposition to proposed legislation based on the transcripts. The paper also provides an overview of sentiment analysis.
An interesting paper by Japanese researchers introduces latent variable models for phrases’ semantic orientations. A word of caution: you have to be in algorithm-reading mode for this paper.
Until the ultimate PR bot is invented, we’ll have to roll up our sleeves and continue reputation management the good old-fashioned way. If you suffer from negativity in the SERPs (define) on searches for your products or services, link building is the answer. Think creatively about how you can get true quality links to gradually move to the top of the pile.
Staying on the subject of words, I want to squeeze in a recommendation. “Letting Go of the Words,” a book by Janice Redish, is a must have for anyone who has anything to do with writing online. With a forward by Steve Krug of “Don’t Make Me Think” fame, it’s an absolutely excellent resource.
Want more search information? ClickZ SEM Archives contain all our search columns, organized by topic.