WOW that is one heck of a question...
I am not certain what you mean in your question so I will answer it to my best understanding.
Any search engine is constructed of (at least) these parts:
1. Collector
2. Indexer
3. Retriever
The collector collects pages (HTML or whatever). The part of AI, to my understanding, is not huge. Just filtering "unreasonable" pages, such as ads etc..
Indexer has no AI as far as I see it. It might have some other interesting algorithms for ranking noun-phrasing etc..
The retriever's AI algorithm comes into fomulating the search terms into complete sentences, the ability to break users sentences into terms and rank that answers based on them.
For example, if the user enters the following question: Who is the smartes man on the face of the earth, the user would expect an answer: Azriel Abramovich (me
) he he he... But no page would hold that answer. You would be happy to see the answer "The smartest man on the face of the earth is Azriel Abramovich" or any other similar text in a page.
So your (easiest) task is to be able to form the structure of the sought answer and then look for it. This is very hard. But it is easiest because you would be happy if the user entered that question. The user will probably enter "man smartes who". And maybe even with typos.
Once you get all the premutations of the text you want to search, you need to be able to cast any sentece which is equivalent to that and check it.
For example: Azriel Abramovich, the smartes man, died today from a too large an
ego..
So, go right ahead. This is a VERY VERY VERY difficult project you wish to undertake.
I wish you luck and success. If you do well you might even make a nice $$$ of it...Not to mention world-wide recognition.
From all the "AI" engines I saw they were not too good and mainly not too fast.
If my answer made ANY sense to you and your question... Then I am happy.
Azriel Abramovich, not so smart, not so modest.
aabramovich@yahoo.com