Using latent semantic analyses and propositionalist methods in text comprehension

In this paper the notion of inference in the area of discourse comprehension and the role of propositions are studied. Two different expository natural texts of high complexity with an implicit inference are presented with different conditions: making the implicit proposition explicit and without any kind of text help. The aims of the study are firstly to see if the differences in the textbase level, the propositional representation, are more pronounced than the differences in the situational model level, as previous literature has suggested; secondly, to see if in the comprehension of complex expository texts a computational method of text analysis such as LSA is able to capture the contribution that the explicitation of implicit inferences usually has been shown in the readability of texts. The results confirm that the presence of an implicit inference is especially helpful, however the LSA values do not show for them a significant improvement in relation to the readability of the texts; concerning the distinction between textbase and situational model, the usual result is confirmed and the main differences are found in the situational model level.