Posted by Carson Ward
If I were to post a restaurant review right here on the SEOmoz blog with the place's name, address, phone number, and a link to its web site, what message would it send? Those of us who have been working in the SEO field for a while might once have immediately responded that citations and links are like a “votes” or “endorsements." However, what if those citations and links are negative – and what if Google can tell the difference?
Suppose I wrote a blog post about a pizza place I visited last weekend, saying that it had the worst pizza in Seattle. It was too expensive, the portion size was too small, and the staff was rude. In the absence of a star rating, does it make sense for Google to count my scathing criticism as just another local citation to boost the restaurant’s placement?
The Name of a Thing Is Not the Thing
For a modern search engine, it is critical to understand both user intent behind a query and the contextual meaning of a word. It must, for example, use contextual clues to determine whether the word “jack” refers to the device I use to change a tire, the common male name, or the slang verb meaning "to steal." Google applies stylometry (the study of language style) in the disambiguation of word and meaning – and they're getting much, much better at it.
Stylometry can also reveal a surprising amount about a person or document. The way that one uses pronouns, for example, could be used to guess whether he or she is telling the truth, prone to depression, or a psychopath with no empathy. On a much more intuitive and basic level, the words we use relay our positive or negative feelings towards a given object.
Google Knows When You’re (Un)Happy
There are dozens of potential uses of stylometry and lexical analysis to improve search quality. Just as Google uses surrounding words to identify what kind of jack I’m talking about, the search engine could also use context to determine whether the adjectives I use in reference to a product, place, or post are positive or negative. Consider the following sentiment analysis patent, filed in 2007 and issued in July of 2011:
Patent No. US 7,987,188 B2 – Domain-Specific Sentiment Classification
“For example the word “small” usually indicates positive sentiment when describing a portable electronic device but can indicate negative sentiment when used to describe the size of a portion served by a restaurant.”
The patent goes on to describe a system that solves this problem by determining what the author is talking about, and then understanding whether the sentiment is positive, negative, or neutral. As with most factors in Google’s algorithm, we can only speculate on the extent to which Google applies this system. Perhaps Rand was on to something similar when he suggested authenticity and passion could be a signal that Google considers.
My sense is that search engines still lack the processing power to analyze sentiment on a large scale. Microsoft’s related patent seems to confirm that (at one point) they felt the same. However, the language of the Google patent and my own experience leads me to believe that Google has been using sentiment analysis on at least local and product results for a while now.
Implications for Local SEO
Over the last several years, the SEO community has proven that citations are like links for local businesses. I will go one step further and suggest that local citations are now analyzed for their sentiment. This allows Google to understand reviews regardless of whether or not the source includes a standard rating scale (e.g. stars). It would follow that positive citations would improve local rankings, while negative citations could harm rankings. To illustrate, consider the following search results:
Note the sentiment-phrases that Google is pulling beneath the places. The Walrus and the Carpenter is ranking quite well, especially considering how far away it is from Seattle’s center. Yes, the reviews are quite positive – but not one of them mentions “steak tartare.”
Could it be that both of these peculiarities result partially from this positive review on The Seattle Times’ Top 10 Restaurants? Is it more than coincidental that the phrase “small plates,” mentioned three times in the linked review, is also listed on the Google Maps/Places pages?
Hundreds of examples prove that Google is picking important sentiments from external websites – not just from review sites like Yelp, but also from sites that don’t necessarily focus on user-generated reviews. Further monitoring over time may be required to properly assess impact on local rankings. It seems clear to me, though, that Google would not go to the effort of determining whether a sentiment was positive or negative and then apply for a patent unless the information is being used in some way.
Getting a site listed in relevant local directories has definite value, but what really matters in improving your local rankings may be sentiment that reviewers express. Furthermore, evidence suggests Google no longer needs an explicit rating to understand how a user feels.
The Future of Links, Brands, and SEO
I doubt very much that even Google has the server power required to perform resource-intensive sentiment analysis on every link – yet. Google’s feeling decoder could easily be tweaked to understand sentiment related to links, domains, brands, articles, and more. Consider the evidence here, Google’s ever-growing powers of lexical understanding, and the potential to improve search quality; the influence of opinion and feeling on search results appears inevitable.
Building a positive relationship is often more difficult than creating negative link bait, but positive relationships are almost always more valuable. While trolling for links might be fun for some, the harsh responses may end up hurting more than feelings. Those concerned with future-proof SEO techniques would be wise to avoid references with negative context or sentiment and instead seek real and positive coverage.
