Natural Language Parsing

This is related to the totally random computer science audit I'm taking for fun this semester, since goodness knows I don't have the patience or the discipline to program with any kind of deadline involved. Anyway, the topic of the week is parsing and stemming with finite-state automata. However, to protect the copyright of the company who wrote it, the book is neccessarily vague about actually using FSAs to analyze words morphologically (rather than just applying the Porter stemmer and so forth). What I want to know, is, has anyone found a way to do this, and also, I am looking for a way to use finite-state automata to build a program that generates langauge rather than parses it - that is, you tell the program, "all consonant clusters must in the form of a fricative followed by a stop" and the program says "OK!" and builds an FSA that represents that. Then you say "and all consonant clusters cannot occur at the ends or beginnings of words" and the program says "OK!" and connects tha FSA to the master FSA that determines a word in general. And the same kind of thing with morphology. This would also be useful for building a program to do my Latin homework for me, since Latin is such a regular language.....

But anyway, the only programs I've seen dealing with word generation are like this one: http://www.fantasist.net/wordmaker2.shtml where you have a finite number of variables and some equations for generating legal combinations. The problem I have with this is that you end up stating the same rule multiple times in multiple places, and you don't get the same ability to nest variables into one another (that is, I would like to be able to define a variable of Fricatives and a variable of Stops, and then define a Variable called Cluster which consists of a formula of FricativeStop, and then I can write Cluster in the other equations rather than repeating FricativeStop over and over again).

The reason for this is that I want to make a conlang, and I want to build a program that can create words in a certain phonology, and can inflect and decline words automatically.

I program in C++ usually, but the class is taught in Scheme (another reason for the audit), and based on what I know of it, I think it's a better language for this sort of thing than C++.

This might be a little technical for the language forum, but I don't really know where else I'd put it.