Context-hunting computers?


During an in-class discussion the other day, a question was posited about how a computer program could be made to search a digitized text for its context (or was it subtext?)

This got me thinking about a tool that can be easily accessed on the internet: the linguist's search engine. Basically, it is a program that can break down a phrase/sentence into its syntactic parts. So, if you enter the statement: "The Chinese community was blamed for the 2003 SARS pandemic", the program will identify that the subject of the statment is the "chinese community".

What I think would be pretty awesome is a tool that can take a search query (like the one stated above) and first identify the syntactic role of the various word-strings. Following this, it could create an ad hoc storage base not only of the separate word-strings, but also for word-strings semantically related (re:synonyms (probably the first three entries of an on-line dictionary)) to those data pieces. This way, you could have a program that can perform a broad non-specific search for a particular phrase.

I had always wondered if there was a way to graphically represent people's reactions following a particular event in the news. With a program like this, I believe that it would be possible to find out how many people thought, e.g., that the Michigan Militia (re: T. McVeigh) was responsible for the attacks of Sep 11/2001, before al-qaeda released a video taking responsability.

I'm not too sure how to make a program like this work, or even if it is feasible. BUT, in theory it seems sort of cool.

Any thoughts?