Wednesday, May 13, 2009

A Very Small Value of "Semantics"

I have now encountered several articles from a variety of my RSS feeds on Google's supposed "semantic stance" towards search, which was launched at Searchology 2009. Having now read Tom Krazit's report for the News division of the CNET Web site, I think I am beginning to form some impressions. On the basis of the few examples I have seen, I have decided that I am reminded of an old joke very popular among engineers (and just as unpopular among mathematicians):

π is equal to 3 for very small values of π (or very large values of 3)

One may call Google's new approach to the delivery of search results "semantic" for very small values of "semantic." Thus, the example that Krazit illustrated, involving the delivery of a rating in the display of a search result from Yelp, basically involves an a priori agreement between Google and Yelp:

  1. Google agrees that the rating is the part the search result that the user most wants to see
  2. Yelp agrees to represent the contents of its pages in a way that makes it easy for Google to find that rating

This may be a small step towards honoring Ludwig Wittgenstein's precept that the "life" of a sign (or, in Google's case, a keyword) lies in how that sign is used; but it is unclear that it can be generalized beyond cases in which the use of that sign is represented by nothing more than an a priori agreement to provide another sign (whose relevance cannot help but be highly context-dependent). Put another way, any view of use that is limited to an association with one or more other signs overlooks the extent to which the user of that sign is a motivated agent. Any inference one makes as to how to deliver information about such a sign/keyword must be guided by the motives of the searcher who invoked that keyword. Wittgenstein would have loved to find a viable way to deduce motives from signs, but he was doomed to frustration. So the best he could do was analyze the nature of his frustrations, and in doing so he educated the rest of us in why questions of semantics are a subtly elusive as they are.

Is there a way out of this frustration? If there is, I doubt that it will have anything to do with new ways of packaging search results, such as what Google has now dubbed "Rich Snippets." My guess is that any progress will come from taking a more verb-based approach to search, recognizing that search is a motivated action and trying to satisfying the searcher through some beneficial exchange of communicative actions, rather than just plunking a collection of Rich Snippets in front of the searcher and hoping for the best. This will require reaching beyond the objective world of more sophisticated systems of signs to represent the content of Web pages into the subjective world of the searcher, which is where all motives reside. This is likely to be difficult but not impossible. We tend to understand the motives of others through the conversations we hold with them. Perhaps, if Google were to spend as much time on building models of those conversations as it builds on building metadata models of all the pages on the World Wide Web, they might finally crack the problem of how to deliver a more semantic approach to search and the delivery of search results.

No comments: