RÉSUMÉ AUTOMATIQUE DE TEXTES: PROBLÈMES ET PERSPECTIVES

Abstract

The present paper focusses on the automatic text summarization (AS), the analysis of linguistic problems related to it and the ways to overcome them, as well as on the perspectives of using some natural language processing computer programs. The author carries out a comparative analysis of two AS programs, MSWord2003 and Pertinence Summarizer, for literary, journalistic and scientific texts. The chosen methodology of comparative analysis allows not only to single out the peculiarities and limitations of each program, but also to make some general conclusions about the problems existing in the process of automatic summarization. The analysis of source texts and results of AS presented in the paper is focused on the correlation between the text genre and the process/result of AS. The analysis does not take into account such factors influencing the quality of summary as the length of the original text, the original language, the subject, etc. The primary hypothesis of the study was the assertion that the quality of automatic summarization of a text directly depends on the genre of this text. The obtained results made it possible to confirm this hypothesis and highlight the interdependence between the level of formalism in the text, which can be explained by its genre, and the pertinence of the summary. The conducted research showed that both AS programs are based, first of all, on morphological and, to a lesser extent, on morpho-syntaxic analysis of the source text. Furthermore, the issue of processing the implicit information available in the text, at the semantic and pragmatic level in particular, still seems unresolved. One of the possible ways to overcome this problem is the dynamic summarization of the text, which necessitates broader participation and involvement of the program user in the process of automatic summarization.