A hybrid architecture for text comprehension

O'Brien et al. (1988) reported that readers generated elaborative inferences only when a text contained characteristics (a strong biasing context or a demand sentence) that made it easy to predict the specific inference that a reader would draw, and virtually eliminated the possibility of the inference being discon-firmed. Garrod et al. (1990), however, offered two qualifications to these conclusions. First, the two text characteristics manipulated may have produced different types of elaborative inferencing: a biasing context results in a passive form of elaborative inferencing, involving setting up a context of interpretation, whereas the presence of a demand sentence invites the reader to actively predict a subsequent expression. Secondly, clear evidence for either type of inference will be apparent only with truly anaphoric materials. This work describes how a passive form of elaborative inferencing, reported by Garrod et al, may be implemented in a connectionist manner. We take the connectionist model proposed by Shastri and Ajjanagadde (1993) in order to represent a text in the form of a network. Next we analyze and discuss how an attentional focus could operate with the proposed reasoner system in order to deal with inference control and anaphora resolution (i.e., antecedent activation), during text understanding. Our own suggestion is that the system proposed by Shastri and Ajjanagadde could be extended by incorporating focus in order to apply it to some open problems related to text processing. However, this extension presents implementational problems due to its local character. Distributed models seem to be in a better position to deal with the shift of the attentional states, given that these systems can learn. Particularly, the model proposed by Sun (1994) is close to our aims, and so we will discuss its implications in order to design a cognitive architecture for text comprehension.