Wednesday, March 11, 2009

Personalized search in a context of Centralized v Distributed (concepts)

Just as American Constitutional Law swings on a pendulum from right to left over time, computer processing has harmonically resonated between local centralized processing (where "all" cpu stuff is done on the local machine like a Mainframe) and distributed processing where "all" the processing is done at remote locations all over the place like a PC network.) I use quotes around "all," because each context can be paralellized. I leave a discussion of my preferences in that regard to another day, in order to observe that a similar discussion has been mooted in the abstract context of search engines. [I should note that I do not confuse the activity of search with the maintenance overhead of search that is mechanically a manifestation of the cpu processing discussion.] What then is distributed search? My first impression of the term led me to theorize an application like "Dogpile," that conglomerates many search engines. By considering outside sources, a major portion of the discussion is being defined in terms of "personalized" search, implying, I guess, "agnostic" search as its antonym. In a corporate environment Google is favoring personalized search at the moment, while Yahoo appears to be defaulting on agnostic. Google's personalized search develops on those items individuals search while logged into their Google profile (gmail, blog, calendar, docs and picasa, without claiming an exhaustive list.) This allows that someone who NEVER logs in may maintain an agnostic search adequately, but in fact normal use would pollute the results so it is not practical to maintain the two separately on the one domain. Yahoo has indeed defaulted to agnostic, but we have no confidence that they will maintain an independent corporate view if personalized search enjoys a period of notoriety. What are my personal concerns with personalized search? Drawing on my old concerns with regard to ranking, I suppose that I do not favor it. Consider if you will a city with a large number of restaurants. They are divided up into ethnic foods, and exactly one restaurant is 7 minutes away from work. In each ethic food category I have a "favorite" food item, but it is prepared to my taste in one restaurant, while at others, I order another item because my favorite is not prepared to my taste there. In proximity to my work place, I have a favorite item too, but last month I ate there 3 meals a day for 8 days straight while working overtime on a project. I am currently sick of that item and don't want to look at another plate of it for a while. In future, I expect to revert back. In this context, what happens to me if I am looking for restaurants on a daily basis using personalized search, because I forget every day where I want to eat for some reason? The results are intuitively limiting, and do not give me adequate choice. As such, I proceed to favor agnostic search. As a service, the choice of attempting to use personalized search is a good marketing gambit, but I don't want Google to go "believing in it like a religion." I am biased in favor of this evil empire over others due to brand loyalty, a practice that the capitalist system does not fail to penalize at every opportunity.

1 comment:

  1. That highlights a problem with personalized search, it's difficult to do knowledge discovery when search engines think you like something. This is also a problem with some personalization algorithms.

    ReplyDelete