Friday, July 21, 2006

search

http://www.readwriteweb.com/archives/search_20_vs_tr.php

I was just reading this article about "Search 2.0". What I think we need to improve search is deeper syntax in the search box. Right now the best we have is boolean queries ("this AND that", "this OR that"). A next step could be syntactic categories. For example,
1. Encyclopedic (what is a thing, or how does a process work)
2. Personal (information about any person: their blogs, photos, etc)
3. Specific data (how many civilians have died in Iraq)
4. Realtime data (weather, game scores, etc)
5. Data sets for living (bus routes, maps, showtimes, etc)
6. Meta (information about the system)

Some of these categories overlap, and each one could be broken down. Once we figure out the best syntactic abstractions, they could be combined into a rich grammar. For example, "realtime metadata relevant to a particular person" (e.g. what types of searches does that person make).

The search system should also perform some data analysis. For example, suppose I had the question, "What are the major theories about JFK's assassination; what kind of people subscribe to these theories; what news articles have been written about the theories; what proportion of magazines have run these articles; and what is the correlation between readership of a magazine and the number of articles?". The search system could dynamically compute the answers to these questions, based on the built-in syntactic concepts.

No comments: