The recent news that Google has been granted a patent on “System and method for supporting editorial opinion in the ranking of search results ” has been taken generally as a sign that Google may delve deeper into the world of social search, which was a hot topic of conversation at the recent Search Engine Strategies Conference (the one I missed due to my shoulder surgery).
But what I find interesting about it is the core: editorial opinion. At some point, algorithms have mothers, and those mothers have opinions. In the related art section, the patent application notes that Ask uses real editors to help it determine results, but that those humans don’t scale:
AskJeeves (www.ask.com) generalizes the application of editorial opinion to a collection of pages. Their editors identify a set of pages that share a common theme (e.g., home pages of airports) and associate this set of pages with specific trigger words (e.g., the word “airport”). When one of the trigger words appears in the query, they present the user with a concise representation of the associated set of pages, allowing the user to choose one. Again, the scope of this technique is restricted to the set of pages that were reviewed by the editors, which tends to be many orders of magnitude smaller than the set of useful pages on the World Wide Web.
Yet this patent very clearly keeps the door wide open for samesaid humans. Quoting from the patent again:
Query themes refer to topics found to be commonly occurring in search queries deployed by users in the network … Editors may, in an implementation consistent with the present invention, develop these query themes based on an examination of search query logs and determining categories of information for which people are entering queries. Exemplary query themes may be “sites that provide free software downloads” or “sites that help people find an accommodation.”
Also critical to the patent is the development of “favored” and “non-favored” sources of information and “editorial opinion parameters”: “the editorial opinion parameter causes the rank of those objects corresponding to favored sources to be increased and a rank of those objects corresponding to non-favored sources to be decreased.”
In short, this is a patent for an algorithm of editorial judgement. It turns on human input, and will, if implemented, but tuned by humans as its shortcomings are exposed.