Descriptive and Theoretical Syntax in the Lexicon
Annie Zaenen, Elisabet Engdahl · 1994
Abstract Many of the computational natural language systems that are built are to some extent based on linguistic frameworks. This is reasonable because in that way they can profit from the research done in those frameworks but also and more importantly because it is necessary to have some framework to organize the data in a domain as large as that of describing a natural language. It is however seen as a drawback of theory-based approaches that they will necessarily be mutually incompatible since there is not one linguistic framework that different researchers agree upon. Of course approaches that lack a theoretical foundation would not necessarily overcome this problem. One might argue that they allow for even more dangerous forms of incompatibility. Where it is in principle possible to predict some of the differences between theoretically based frameworks on the basis of their principles, there would be no clue to guide a comparison between descriptions that are based on no principles at all (assuming that such could exist).