8. On the notions of theme and topic in psychological process models of text comprehension

Latent semantic analysis (LSA) is used to define the theme of a text and to generate summaries automatically. The theme information – the already known information-in a text can be represented as a vector in semantic space; the text provides new information about this theme, potentially modifying and expanding the semantic space itself. Vectors can similarly represent subsections of a text. LSA can be used to select from each subsection the most typical and most important sentence, thus generating a kind of summary automatically. latent semantic analysis. 2 Distinctions like given-new, topic-comment, or theme-rheme have played an important role in psycholinguistic investigations of sentence processing (e.g., Havilland & Clark, 1974; Gernsbacher, 1990). On the other hand, theme and topic are not basic concepts in psychological models of text comprehension, at least not explicitly. However, certain features in these models correspond, not always directly, to the notions of theme and topic. I shall describe these features here as a first step in exploring the analogies and correspondences between psychological process models of language understanding and the linguistic notions of theme and topic. I shall focus here on a specific process model of text comprehension, the for a broader discussion of the psychological literature in this area see Gernsbacher, 1994). Only those features of the construction-integration theory directly relevant to the notions of topic and theme will be discussed here. The construction-integration model of text comprehension attempts to simulate the computations involved in the construction of a mental representation of a text in human comprehension. Fundamental to the theory is a distinction between different aspects of the mental representation of a text. The textbase comprises that part of the mental representation that is directly derived from the text. Naturally, a great deal of knowledge on the part of the comprehender is necessary to construct a textbase, both linguistic knowledge and general world knowledge, but this is knowledge employed in the service of understanding what is directly contained in the text. The mental representation of a text, however, comprises more than this textbase-the comprehender's prior knowledge is used to fill in missing links in the text and to elaborate what the text actually says. Thus, the final product of comprehension, the situation model , is a mixture of structures that are derived directly form the text and structures that have been added from prior knowledge. Just what this mixture will be depends on …