Letter to the Editor: On the Need for Careful Description of NL Prototypes

As a result of much computat ion linguistic research in the 1970's, natural language (NL) processors have begun to rival more formal means of computer access in some instances. Unfortunately, very little is known, even by applied NL researchers themselves, about the true capabilities of existing systems. This is partly due to the fact that most l i terature, at least in journals such as this one, has focused on technical or theoretical questions, paying less at tention to practical matters. We believe the state of the art in NL system design allows authors to become more serious about matters of usability, without losing sight of the theoretical basis on which various facilities are being designed. Many researchers have observed the usefulness of obtaining actual " h a n d s o n " experience with an NL system in order to evaluate it or even truly understand its workings (e.g., Ballard 1981, Harris 1980, Morris 1979, Petrick 1976, Sondheimer 1978, Tennant 1979). This suggests to us that a profitable and responsible way to organize written reports of prototype NL systems is to attempt to "s t imulate" a one-on-one demonstration. Of course, certain aspects of a given system will deserve separate papers but, even in those cases, it is important to understand their role in the overall processing. We now propose a standard format for descriptions of pro to type or near -pro to type natural language processors.