Skip to Main content Skip to Navigation
Theses

Compréhension de textes et représentation des relations causales

Abstract : This work is about the notion of causal distance in the mental representation of read text (hypothesis of situation model). I reviewed the literature (including philosophical one) so as to know how to define causality in reality. Such a definition seeming out of reach, I developed operational definitions of causal distance, both in reality and in the situation model.
I ran two experiments on more or less complete causal chains (that is, sequences of sentences in which each describes the consequence of the previous one) taken for vulgarization texts, in which I collected likeliness judgements on cause-consequence couples. It shows that mental causal distance (likeliness) increases as intermediate elements are dropped (approximation of real causal distance). There is no effect of the reading rank on likeliness. Also, likeliness grows with the familiarity of the participant with the theme of the sequence. Descriptors of time, space and protagonists are proposed, which allows a finer description of the relation between events described by two sentences, rather than the descriptors employed in Zwaan's multiple indexing model. Those descriptors were valued for all cause-consequence couples tested, and they are predictors (41% of explained variance) of likeliness judgements. Also, I identified some descriptors indicating typical causal relation, and these are not predictors. Finally, causality seems to be a dimension of the situation model, but it can be predicted by other situational dimensions like time, space, and protagonists. We suggest that these situational data can be the basis for a raw decision process for carrying out a causal inference, while reading a text.
Document type :
Theses
Complete list of metadatas

Cited literature [225 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-00161089
Contributor : Amal Guha <>
Submitted on : Monday, July 9, 2007 - 5:27:31 PM
Last modification on : Wednesday, October 14, 2020 - 3:41:42 AM
Long-term archiving on: : Thursday, April 8, 2010 - 10:46:41 PM

Identifiers

  • HAL Id : tel-00161089, version 1

Collections

Citation

Amal Guha. Compréhension de textes et représentation des relations causales. Linguistique. Université Paris Sud - Paris XI, 2003. Français. ⟨tel-00161089⟩

Share

Metrics

Record views

504

Files downloads

11500