A multidimensional dynamic framework for handling simple interruption phenomena, anaphoric pronouns and definite descriptions

The candidate connrms that the work submitted is his own and that appropriate credit has been given where reference has been made to the work of others. 2 1 Abstract The last two decades have witnessed the development of theories dealing with discourse. Contributions from Linguistics, Philosophy, and Logics have helped to expand the knowledge of the eld. But, independently of the researchers' viewpoint, all theories make the same basic claim, namely: discourse is a complex structured abstract entity. If a theory is to succeed it must take into account such structure. Therefore, questions such as \What structures do discourses have?", \How many structure sorts can be assigned to discourses?", \Can discourse structure sorts be uniied?", and the like, must be be asked (and answered). This thesis proposes a formal semantical model as an answer to the rst question. This is not a work in Linguistics; it, however, is based on a series of well established results from Linguistics research. For instance, it presupposes that discourses can be hierarchically organized having discourse segments, or discourse blocks, as their basic components. Also, it presupposes that coordination and subordination are the basic relations among discourse blocks. However, diierently from the literature where the hierarchical organization corresponds to the well known tree structure, leading in one way or another to stack oriented processing, this work proposes a \list" oriented processing. The stack oriented processing has been vindicated in the literature as a tool for processing discourses. Among many other things, the stack-block structure allow us to explain how anaphoric relations are possible even across segment borders. This would be ne for continuous discourses. But, it clearly fails to account for interruptions. Sometimes, a discourse can be interrupted and resumed later on. A simple example of this phenomenon occurs when someone introduces a person to a group and discovers that that person is not so well-known as (s)he thought. Trying to recover the situation, a parenthetical background explanation would be given; after that the discourse might be resumed naturally. But, if the resumption is almost always taken in a backward direction (and therefore the stack model will work ne), there might be the case that a \forward" move 2 into the parenthesis is needed. As a consequence the stack model ought to be relaxed (the stack is not a stack, after all) or hardened (stacks of stacks, and the stack-like have been prescribed for …