Does Google use humor as a search algorithm signal? My first instinct is to state, without hesitation, that there is no humor signal. There’s no analysis of puns or witticisms or double-entendres or other grammatical humor.
The clever pun you use as your browser title will be lost on Googlebot and, truth be told, the majority of your readers as well. Googlebot doesn’t have a funny bone.
Yet the algorithm will reward humor through other signals. Something that’s really funny will attract links and generate a high number of Likes, Tweets and other social gestures.
So humor is reflected in the algorithm through the link graph and social signals.
But that’s an indirect signal, sort of like a backhanded compliment. Google doesn’t really know it’s funny, it just knows it’s popular and engaging.
Google might be able to determine whether a document is humorous or not. Computational humor, a relatively new field of study, has shown a moderate amount of promise.
Early computational humor research focused on producing jokes, resulting in the Joke Analysis and Production Engine (JAPE) and HAHAcronym, which does just what it says. More recently, there’s been a greater emphasis on identifying humor using natural language processing and machine learning (pdf).
Researchers were able to train an algorithm, with high accuracy, to identify humorous one-liners from standard headlines and even proverbs, which had similar sentence construction.
Is Google interested in humor? It’s tough to tell. But a recent patent, brought to my attention by (who else) Bill Slawski, titled Embedded Communication of Link Information contains an interesting idea.
The improved link tags may allow the publishers to communicate additional information, such as opinions, about the content locations and/or document locations. The additional information may be along one or more dimensions. Therefore, different information may be conveyed at the same time. For example, one dimension may indicate that a content location and/or a document location is offensive as well as funny.
For example, “offensive=very” or “funny=somewhat.”
Coincidence? Probably. Nevertheless, I think it’s interesting that humor was one of the examples selected. And most of the Googlers I know have a keen wit.
Humor and Quality
At the end of the day the real question is whether humor and quality are related? Does the incidence of certain types of puns or other humor signal quality? Could quality be predicted based on Monty Python references? Or the presence of alliteration?
Could you construct a database of humor? Monty Python phrases. Caddy Shack quotes. Famous puns. Palindromes. Internet memes. Knock knock jokes. Using that database, could each type of humor be given a score based on how often it appears in what quality of content?
Those scores in place, you can begin to use humor as a signal to help ‘turn up’ high quality content. Perhaps … to 11.
Perhaps the relationship between content quality and humor doesn’t exist. There’s a tremendous amount of personal taste involved and the cultural differences are also substantial. In addition, much of our humor is visual in nature and for all the advances in image search Googlebot still can’t really see.
Yet, humor is clearly a compelling part of the content puzzle. And there’s a part of me that believes that the mere presence of humor indicates a higher level of care and passion, which dovetails nicely with Google’s recent guidance on high quality sites. I doubt you’re going to see humor (at least intentional humor) on splogs or MFA sites.
I still don’t think there’s a humor signal. But maybe there should be. It would be nice to reward writers for their spelling, grammar and creativity.