Super Personalized Search

Search engines, e.g. Google, have transformed to personalized search engines, i.e. they are learning your habits of search and can nicely predict what you are going to search, and not only what you are searching now. These algorithms usually depend on your previous search history and other smart things that can be learned from other people's searches.

I want to suggest to expand this to new dimensions (probably Google already does this, but I don't know of it J). The first rather trivial dimension is time. There is usually a temporal dynamics of the searches, i.e. if you search X there is larger probability that you'll search Y only within a narrow time window, e.g. one day. This statistics can be learned from other users. I've heard of research done on search patterns of medication and a time-lapse of search of side-effect as a means to detect side-effects via searches. In other words, if I search for drug X and a month later I search for "head-aches", if enough people repeat this search, perhaps X's side effect is a head-ache after a month. To conclude this dimension, the search should also take into consideration delays in search and narrow time windows of X-Y search correlations.

Another cute dimension to super personalize the search is outside events, e.g. weather. The search can correlate between your own personal attitudes toward weather phenomena, e.g. snow, heat-waves, etc. Thus, taken from other sites the weather at your own location, the search can be refined when you search for "coffee shops": if it's snowing it will direct you to a more hot-coffee oriented shop, whereas if it's hot, it will direct you to ice-coffee shops.

Other local and/or global events can also shape your personalized searches, as specific people react differently to events, e.g. election, holidays, etc. Correlating your search patterns on similar events and cross-referencing it with people within your search-cluster can further refine the search results when an event occur.

Finally, the search can be refined to your own current state of affairs. By accessing your own social media, e.g. Facebook, tweeter, it can refine the search for your own and your friends state. For example, if you just posted "This is not my day", your search can be refined to happier sites so as to lighten up your day. Another example is that if a friend of yours posted that she seeks something, your own searches can be refined so as to hint to a solution to your friend's problem, even though you searched for something else. This can augment both your social interaction media, as well as your search results.

Once along these lines, one can possibly think of other ways to super-personalize the search. I encourage you to think of them, implement them, open a start-up, be bought by Google and become a millionaire. A small (non-committing) comment in the blog will be appreciated.