A system for text and content analysis

Abstract

THE INTRODUCTION of the computer in the last ten years as a tool for the exploration and analysis of literary data has not been a substitute for the intelligence of the researcher but more of an extension of his memory and organizing capabilities. In the investigation of many disciplines based on textual data (literature, history, sociology, philosophy, theology), the researcher often has to manage a vast quantity of information. As he must frequently parse, recall or organize his text and make inferences from it, he may write cards, underline, annotate with numbers, copy references, generate new texts, produce concept associations, build concordances and so on. In fact, many of his procedures treat the text as nothing more than a set of elements held together by an organizing structure. At this physical level, a computer can be of great assistance, for it can offer many systematically organized data structures for a text and recall hierarchically organized information that varies according to the chosen level of interpretation, be it morphological, lexical or even conceptual. A textual data processing system for literary and social sciences, called SATO (Systime d'analyse des textes par ordinateur), has been implemented with a view of providing an interaction between textual data description and its interpretation. It helps to generate hypotheses which serve as starting points for the researcher's interpretations. The system, conceived and written at the Universit6 du Qu6bec in Montreal, has been working since 1973. The first part is presented here; a second part is now under development, involving scholars from philosophy, linguistics and computer science. The SATO system is related to similar systems in the fields of literary textual data processing and content analysis, such as the General Inquirer.' In the first these, many textpack systems oriented towards the manipulation of large textual data bases have produced indexes, concordances and basic frequency counts. Used mainly for stylistic, philological and lexical studies, as well as the statistical analysis of authorship, these textpacks are usually programmed to present a simple interface to the literary scholar who has no special knowledge of computer programming. The second of these areas, content analysis, has been more concerned with methodology. Of secondary im ortance is the occasionally large volume of textual data, which may come from newspapers, interviews, etc. The analysis techniques are more sophisticated, for example cluster analysis, factorial analysis, categorial and correlation analysis. The SATO system has been conceived with these two functions in mind, i.e., classical literary research and content analysis either on standard or transformed texts. Like the other