INTRODUCTION

The computer is a problem-solving tool, and the use of computing equipment in linguistics has brought about an increased emphasis on problem-solving. Consequently, the more the research takes on a problem-solving aspect, the more the use of computers becomes meaningful.

It is useful, therefore, to give some thought to the nature of linguistic problems and to the methods and tools required for their solution, and in this context, to assess the application of the computer as a problem-solving tool in linguistics.

Most usually, linguistic problems are organized on the basis of the conventional divisions of the field of linguistics such as the best known of these, into historical and descriptive linguistics. In historical linguistics then, problems are those of comparison and reconstruction, etymology, historical dialectology, and the like. A problem can here most often be defined in terms of the description or interpretation of a given historical process, or of the origin and historical background of a particular phenomenon.

In descriptive linguistics, problems are most often classified in terms of the different levels of description that a particular linguistic researcher accepts. Within the well-known division of phonemic, morphological, and syntactic problems, an individual problem will be concerned with the description or interpretation of a given aspect of phonology, morphology, or syntax in a particular language.

In both major branches of linguistics, the formulation of a prob lem will of course be strongly controlled by the particular theoretical frame of reference that the researcher adhers to. Thus, in historical linguistics, his frame of reference may either be comparative in the older tradition, or structuralist. In descriptive linguistics, the older descriptivist or one of the more recent approaches such as transformational theory may be chosen.

A further area of linguistic problem-solving is that of theory itself. Here, the problem is not posed primarily in terms of a given set of data that are to be described and/or interpreted, but in terms of a theoretical position that has to be further developed and applied to a body of data. In most theoretical work, the problems turn out to be more philosophical than operational in nature, since the question here is one of greater depth of explanation.

With the introduction of the use of computing equipment into linguistics, it has become possible to describe and categorize linguistic problems in less conventional terms. Instead of the subdivisions of the field of linguistics, the nature of the process required for the solution of a problem may be used to define the the problem. This new type of categorization of problems, based on the problem-solving process, is necessary if one is to assess properly the role of the computer as a problem-solving tool in any discipline.

Two important characteristics of computing equipment can aid in such an assessment. On the one hand, the memory capacity of modern computers, as well as their speed of operation, make possible the processing of extremely large bodies of data. Data of great bulk can now be processed within a finite span of time whereas earlier this analysis would have required decades of work by human researchers. The second important characteristic of computing equipment is its mechanical logic, that is, the well-known fact that data-processing machines carry out every single instruction to the letter. They lack the human capacity for overriding instructions spontaneously, a capacity which makes human beings both more original and less reliable than computing equipment.

The first-named characteristic of computers can be called their data-processing capability, the second, their logical capability. The data-processing function and the logical function can be used as the two poles of a continuum on which the nature of linguistic problems and their solution with the aid of computing equipment can be discussed. I have written on this subject on a previous occasion under the heading of ‘Computer Participation in Linguistic Research’, (Lg. 38. 385-9 [1962]). My basic point concerned the varying degree to which computation is an intrinsic part of the linguistic research process in which it is applied. Three fundamental degrees were posited there: ‘language data collection, which is essentially a form of bookkeeping; computer programs using the results of linguistic research; and automation of linguistic research procedures’(p. 385). These degrees can be placed along the above-mentioned continuum. The lowest degree relies almost entirely on the data-processing capability, the highest degree almost entirely on the logical capability. The criterion is the extent to which the computer program ‘exercises’ linguistic ‘judgment’. ‘The “judgment” of a computer program thus consists in the increasing diversity of conditions which it is capable of taking into account in making its string of yes/no decisions’ (p. 389).

It is possible to take still another view of this question of the ‘judgment’ exercised by the computer, namely, in terms of the stage in the application of the equipment at which the fundamental research decisions are made. Are the decisions made in the manual preparation of the input for processing by the computer? Are the decisions made in the program itself? Are the basic decisions made in the interpretation and further consideration by humans of the output?

It is clear that the work of the computer is most closely associated with the research process when the basic research decisions are written into the program. That is, the researcher uses the computer as an extension of his intellect rather than as an accelerator of his filing cabinet. The logical function of the computer manifests itself here in giving back to the researcher the exact consequence of his assumptions. This allows him to verify, with a degree of precision not achievable without the use of computing equipment, the exact consequences of his assumptions. It is for this reason that this highest degree of computer participation has great theoretical as well as practical significance for the field of linguistics.

The papers in this volume exemplify various degrees of computer participation in several fields of linguistics. The book, organized basically in terms of the two fundamental capabilities of the computer, falls into two categories: data-processing problems and systems problems (i. e. problems of logical design). Within each of these categories, the further classification of the contributions follows the more conventional subdivisions of the field of linguistics.

Paul L. Garvin

The Bunker-Ramo Corporation

Canoga Park, California

November, 1965