The Core Language Engine

Yes, a lot has happened since this book was written--statistical grammars, grammar induction, constraint-based grammars, not to mention that whole web thing. No matter. This book is still a valuable stepping stone on the path to NLP nirvana, and everyone who aspires to do research in NLP should come to grips with it. Why? Because it presents what is really the example par excellance of a unification-based grammar, the core language engine. This is really the first attempt which had any sucess at all to "scale up" nlp techniques to create a system which isn't just another toy. As such, it is one of a very very few descriptions of what it is like to build a BIG program in a logic programming language. Is this a book which points the way to future reasarch? Well, the authors have pretty much "exhausted the search space" of unification grammars here. Both their potential and their painful limitations have been plummed, and are on display here. But even though the time of unification grammars has passed, the lessons learned in the construction of the Core Language Engine represent, IMHO, a signature achivement, and real ground gained towards the goal. Moreover, the spirit of this research lives on. For a bang-up-to-date discussion of where this stream of research has lead, check out Blackburn & Bos's book "Representation and Inference in Natural Language".