Fake it in the most convincing way possible. If you're only handling queries, there's no reason to make a full blown NLP system. Queries have a certain form... "who are..." "where are...", etc. So come up with a few simple heuristics for handling these queries then map them to SQL or whatever...
Good luck.
Most important is the way to interact.
Start collecting realistic queries.
Clarify the queries you a are making: formulate them in NL.
This way you are helping people to formulate their own queries (which is not easy).
Allow incremental queries, precisizing the previous query.
Use an SQL near algebra, not necessarily SQL itself.
Log things, ask for evaluations (did not help me/...).
Background reading material which at least is illuminating is:
http://citeseer.ist.psu.edu/
"Studying the Language and Structure in
Non-Programmers?Solutions to Programming
Problems"
by John F. Pane, Chotirat 揂nn?Ratanamahatana and Brad A. Myers
Hello I hear you need to write a NLP.
1. Learn to write a parser or use a ready made parser already out there. Its not as hard as it might seem for simple tasks. Investigate Standard ML parsers
2. Learn the grammatical rules of the language you are writing it for. There are a lot of researchers in Europe ( especially France, don't ask why?) that are really good at this stuff I have seen some tools made by them in Java.
You might find a good Finite Automaton algorithm to help you here in one of their papers. Research should be centred around "phrase utterance" , "text summation", "intelligent agents" and other stuff that I cannot remember now but can tell you if you give me your e-mail.
3. Also it depends on the level of processing you are doing and the level of intelligence you require. I think you need to specifically do something more inteligent then look into machine learning techniques GOOGLE will help here!
4. Some people make these tools from ready made APIs AIML is one such example. Its based on XML. Other systems include expert analysis systems where you can search for key words. Look into systems like eliza and the history behind that. and the evolution behind similar AI systems.
5. Start simple and then build up, this sounds like a University Project and after doing several of these "Seeming Impossible" I can assure you it can be done . So have a positive attitude and work hard!
When I am a bit more free I will try to give you something a bit more helpful.
BUT don't forget to put some of your own effort in too.
best of luck I hope you are successful
Prashant