I propose a model for determining the hearer's attentional state which depends solely on a list of salient discourse entities (S-list). The ordering among the elements of the S-list covers also the function of the backward-looking center in the cen- tering model. The ranking criteria for the S-list are based on the distinction between hearer-old and hearer-new discourse entities and incorporate preferences for inter- and intra-sentential anaphora. The model is the basis for an algorithm which operates incrementally, word by word.

this paper gives a comprehensive picture of a complex, yet not explicitly spelled-out theory of discourse coherence, the centering model (Grosz, Joshi, and Weinstein, 1983, 1995) marked a major step in clarifying the relationship between attentional states and (local) discourse segment structure. More precisely, the centering model accounts for the interactions between local coherence and preferential choices of referring expressions. It relates differences in coherence (in part) to varying demands on inferences as required by different types of referring expressions, given a particular attentional state of the hearer in a discourse setting (Grosz, Joshi, and Weinstein 1995, 204-205). The claim is made then that the lower the inference load put on the hearer, the more coherent the underlying discourse appears. The centering model as formulated by Grosz, Joshi, and Weinstein (1995) refines the structure of "centers" of discourse, which are conceived as the representational device for the attentional state at the local level of discourse. They distinguish two basic types of centers, which can be assigned to each utterance Ui--a single backward- looking center, Cb(Ui), and a partially ordered set of discourse entities, the forward- looking centers, Cf(Ui). The ordering on Cf is relevant for determining the Cb. It can be viewed as a salience ranking that reflects the assumption that the higher the ranking of a discourse entity in Cf, the more likely it will be mentioned again in the immediately following utterance. Thus, given an adequate ordering of the discourse entities in Cf, the costs of computations necessary to establish local coherence are minimized

From the analysis of naturally occurring texts we obtained evidence for the systematic interaction between nominal anaphora and metonymies. This leads us to postulate an integrated model incorporating both phenomena simultaneously. The consideration of discourse constraints for metonymy resolution allows us to challenge the commonly held view that the interpretation of metonymies should proceed from a literal-meaning-first approach. Thus, we argue for an equally balanced treatment of literal and figurative language use.

We specify an algorithm that builds up a hierarchy of referential discourse segments from local centering data. The spatial extension and nesting of these discourse segments constrain the reachability of potential antecedents of an anaphoric expression beyond the local level of adjacent center pairs. Thus, the centering model is scaled up to the level of the global referential structure of discourse. An empirical evaluation of the algorithm is supplied.

Abstract not found

Abstract not found

this paper, we shall challenge this view. We stipulate that medical texts, as any other text sort, exhibit textual structures and that disregarding these structural relations will lead to underdetermined or even invalid content representations. To render support to our argument we conducted an empirical investigation of medical findings reports from a large clinical text database in order to assess whether this issue is really relevant. First, we will describe the experimental setting and elaborate on the quantitative distribution of various text phenomena in the sample. We will then turn to the consequences of not taking textual structures into account and show how referentially incoherent and referentially invalid text knowledge representation structures are likely to emerge. This is illustrated considering a small text fragment as analyzed by our system prototype, the Medical Knowledge SYNDIKATE. We shall focus on those aspects of the system design which account for the proper analysis of text phenomena. Up to now, such a functionality has to the best of our knowledge not been provided by any other system for medical text processing (for a survey, cf. [4]).

As a response to the requirements imposed by real-world natural language processing, we argue for a design of natural language grammars and their associated parsers in which declarative knowledge about language structure and procedural knowledge about language use are equally balanced within a strictly object-oriented specication and implementation framework. In particular, we introduce fundamental messagepassing protocols for object-oriented parsing, which include, besides the one for basic dependency parsing, protocols for ambiguity handling, robustness, backtracking, preferential and predictive parsing, as well as textual reference resolution. Based on an empirical evaluation, we also provide reasons for sacricing completeness of the parse in favor of eOEciency gains. 1 2 1. INTRODUCTION Over the past decades the design of natural language grammars and their parsers was almost entirely based on competence considerations (Chomsky, 1965). These hailed pure declarativism (Shieber,...

This thesis concerns the notion of `information structure': informally, organization of information in an utterance with respect to the context. Information structure has been recognized as a critical element in a number of computer applications: e.g., selection of contextually appropriate forms in machine translation and speech generation, and analysis of text readability in computer-assisted writing systems. One of the problems involved in these applications is how to identify information structure in extended texts. This problem is often ignored, assumed to be trivial, or reduced to a sub-problem that does not correspond to the complexity of realistic texts. A handful of computational proposals face the problem directly, but they are generally limited in coverage and all suffer from lack of evaluation. To fully demonstrate ...

We argue for a performance-based design of natural language grammars and their associated parsers in order to meet the constraints imposed by real-world NLP. Our approach incorporates declarative and procedural knowledge about language and language use within an object-oriented specification framework. We discuss several messagepassing protocols for parsing and provide reasons for sacrificing completeness of the parse in favor of efficiency based on a preliminary empirical evaluation. 1