Emotion is one of the more interesting elements of Google’s search engine algorithm, at least in my opinion. It is unclear if Google’s algorithm takes emotion into account when determining search rankings, but a case can be made for either. Traditional SEO belief was that the more links into a site, the better. Google has made it their goal to make their algorithm more “human;” it should view page relevance to a query the same way a human would. If that’s the case, than emotion should be a factor. If I wrote a blog post talking about an awful Chinese restaurant that I went to and in this post I criticized the restaurant throughout, a human ranker would move that restaurant down in a list of local Chinese restaurants.
A blog post from SEO Moz believes that positive or negative sentiment is a factor in Google’s algorithm. The algorithm already considers context when reviewing a query that can have multiple meanings. The article calls this “stylometry,” the study of language style. If Google can already use context to determine the correct meaning of the word, what’s not to say that they can’t use context to determine positive or negative sentiment? If you dig a little deeper, you can find that this is something that Google is definitely pursuing. In April 2007, Google filed for a patent titled “Large-scale sentiment analysis” and the patent was issued to Google in August 2011. The patent abstract reads, “A method for determining a sentiment associated with an entity includes inputting a plurality of texts associated with the entity, labeling seed words in the plurality of texts as positive or negative, determining a score estimate for the plurality of words based on the labeling…” and eventually getting a corresponding sentiment score.
So Google definitely uses sentiment analysis, right? Not necessarily. The issue of sentiment analysis came to a head just over a year ago when the New York Times reported a disturbing story about an online retailed called DecorMyEyes. In short, one company treated its customers horribly, even threatening and terrorizing them, on the premise that complaints and negative reviews posted online created links, helping search engine optimization. This went beyond bad reviews; the head of the company created several fake personas, overcharged and threatened customers, even telling the employer of one customer that their employee was a homosexual drug dealer. But the head of the company was unremorseful because the worse he acted, the better business got. This story of course raised the idea that Google needed to include more about sentiment in its search rankings.
However, in the interest of providing users with the most relevant pages and best information related to their query, penalizing all links in the context of negative sentiment is not always effective. For example, let’s take something that I hate, the New York Yankees (and there are plenty of Red Sox fans to back me up on this). If I write a blog post about how much I can’t stand the Yankees and link it to their page, this should not hurt the Yankees’ search rankings. The Yankees official home page should not be dropped below a team store with a lot of good reviews because of negative sentiment.
Google responded to the DecorMyEyes story with an official blog post in which they claimed they would not be enacting sentiment analysis across the board for this reason. The example they gave was elected officials (probably a more important example than a baseball team). Politicians often have people strongly for or against them, and the links that criticize them should not affect them in the search rankings. While a lot of people may disagree with the policies of Rick Santorum, his official page and information about his policies should be high results on search engines (I chose Santorum because he has his own Google problem. Google him if you don’t know what I’m talking about).
The truth of course is that sentiment analysis has its merits and drawbacks. I believe that there are particular cases in which sentiment analysis could be very beneficial, like if the search query is for a commodity, restaurant or retailer. I would not be surprised to see Google use sentiment analysis with Google Places by integrating the reviews of users. However, with more informative queries, implementing sentiment analysis may be harmful. I think the key for Google is adding something to the algorithm that can distinguish between the two. They may already be doing this, but as Google’s algorithm is always kept top secret, we have no way of definitely knowing. If Google does announce any further changes, we will definitely keep you up to date.