IntroductionFor many years, literary scholars, folklorists, anthropologists, and others have been developing a rich pool of analytical techniques for selective critical interpretation of texts. To these have been added, through comparatively recent advances in computer technology, language processing tools such as concordance generators and statistical analysis programs that have freed the humanities scholar from timeconsuming clerical chores, and supported his qualitative judgments with quantitative statements. While such tools allow the humanist to use his time more effectively, they do not, however, in any way augment his intellectual processes. Computer-based approaches to the study of cognition in linguistics, cognitive psychology, and artificial intelligence are currently having an important impact on the study of language and the representation of the underlying knowledge which supports the comprehension process. This new area of research offers approaches to the construction of computer programs which would assist the user in the substance of his task, as opposed to programs that allow him to perform clerical activities more effectively. The general trend is to devise computer programs which model human natural-language comprehension and then delegate some of the user's intellectual functions to the machine.One such function might be the deduction of implicit facts and relations from a text and their explicit representation in a data base. To do this, a computer must first be programmed to "understand" language. Twenty years of experience in the field of natural-language processing have shown that an automated language "understanding" system must incorporate many kinds of knowledge. The fundamental problem facing builders of such systems is how to represent in the computer the many kinds of knowledge which a person uses in understanding natural-language text. Although there has been a certain amount of arm-waving on the subject of representing extralinguistic or "encyclopedic" knowledge since the time before computational linguistics had a name, few of the concepts used appeared at all productive, and even fewer were sufficiently explicated to determine how they might be used. The focus was on syntax, which appeared more tractable than semantics and pragmatics) Recent computational linguistic research in knowledge representation for automated understanding of natural language text seems to constitute a significant departure from the generally fuzzy notions of the past, and in the authors' opinion, has considerable potential for the future. As Benzon and Hays (1976) rightly observed, the methods of language analysis practiced by computational linguists differ both in purpose and in techniques from those of the literary theoretician. However, we believe that the tools and approaches developed by the former can be of interest and utility to the latter. It is for this reason that we offer this paper, which focuses on recent developments in the representation of linguistic and extralinguistic knowledge for aut...

No abstract