Two perspectives to text structure representation

Abstract

This study outlines a two-layered system of text-structure representation for factual English texts. Because the study focuses on factual texts, it defines text as consisting of a group of propositions which relate to each other so that an identifiable discourse topic emerges. For practical purposes text is also described as a textual topic-comment message. Text structure is, accordingly, defined as the set of relations by means of which the propositions of the textual comment are joined to each other and, directly or indirectly, to the discourse topic. The text is assumed to contain both informatively indispensable propositions and those which do not increase information in their context but are included in the text to make it more comprehensible, convincing or attractive to the reader. To account for this textual feature text structure is approached from the two perspectives of message structure and message trategies. The message structure consists of the relations joining the informatively indispensable propositions of the textual comment to each other and, directly or indirectly, to the discourse topic. The message strategies are defined by the relations existing between the indispensable and the dispensable propositions. The message structure shows two kinds of relations: message-core relations and logical relations. The top-hierarchy propositions of the textual comment relate to the discourse topic by means of a dominant type of message-core relation. The various types of message-core relations result from the various illocutionary purposes that the text may show as one speech act by the writer. Logical relations carry the meaning of clause-connecting conjunctions. Various types are distinguished also among message-strategic relations. Thus a number of more specified text-structural functions/relations are derived from each of the above three main categories of text-structural relations. Stereotypical patterns of message structure are then suggested for the various text types. Finally, the model is tested on a corpus of 18 authentic texts/extracts of texts including scientific articles, textbook texts, and press reportage. The two levels of analysis are carried out side by side so that a group of propositions is coded for its message-core function and for the messagestrategic pattern by means of which the message core function is realized.