... this paper, we argue that, to handle explanation dialogues successfully, a discourse model must include information about the intended effect of individual parts of the text on the hearer, as well as how the parts relate to one another rhetorically. We present a text planner that records this information and show how the resulting structure is used to respond appropriately to a follow-up question.

To explain complex phenomena, an explanation system must be able to select information from a formal representation of domain knowledge, organize the selected information into multisentential discourse plans, and realize the discourse plans in text. Although recent years have witnessed significant progress in the development of sophisticated computational mechanisms for explanation, empirical results have been limited. This paper reports on a seven-year effort to empirically study explanation generation from semantically rich, large-scale knowledge bases. In particular, it describes KNIGHT, a robust explanation system that constructs multisentential and multiparagraph explanations from the Biology Knowledge Base, a large-scale knowledge base in the domain of botanical anatomy, physiology, and development. We introduce the Two-Panel evaluation methodology and describe how KNIGHT'S performance was assessed with this methodology in the most extensive empirical evaluation conducted on an explanation system. In this evaluation, KNIGHT scored within "half a grade " of domain experts, and its performance exceeded that of one of the domain experts. 1.

This paper presents the Joyce system as an example of a fully-implemented, application-oriented text generation system. Joyce covers the whole range of tasks associated with text generation, from content selection to morphological processing. It was developped as part of the interface of the software design environment Ulysses. The following design goals were set for it: -- The generated text must be of sufficiently high quality so that the user community of the underlying application accepts it as part of the documentation of software designs...

This paper describes a novel method by which a spoken dialogue system can learn to choose an optimal dialogue strategy from its experience interacting with human users. The method is

Introduction During the past few years we have been concerned with developing models for the automatic planning and realization of report texts within technical sublanguages of English and French. Since 1987 we have been implementing Meaning-Text language models (MTMs) [6, 7] for the task of realizing sentences from semantic specifications that are output by a text planner. A relatively complete MTM implementation for English was tested in tlie domain of operating system audit summaries in the Gossip project of 1987-89 [3]. At COLING-90 a report was given on the fully operational FoG system for generating marine forecasts in botli English and Frencli at wcatlier centres in Eastern Canada [1]. The work reported on here concerns the experimental generation of extended bilingual summaries of Canadian statistical data. Our first focus has been on labour force sur- veys (LFS), where an extensive corpus of published reports in each language is available for empirical study. Tlie current LFS

This paper outlines a number of problems with RST's elaboration relation, and discusses a new model of text structure that results from leaving this relation out of the set of relations. In this model, trees of interclausal/intersentential relations account for the local coherence of a text, while its global coherence is accounted for by a separate device: global focus. 1 Introduction Many theories of discourse propose that a coherent text is one whose clauses, sentences and text spans (or perhaps the propositions expressed by these text units) stand in particular relations to one another. The basic motivation in these theories stems from the observation that a text is more than a sequence of independent units: whether a particular unit makes sense in a given discourse depends not only on this unit by itself, but also on its relationship with the other units in the discourse. This claim has been spelled out in many dierent ways, but there are two requirements that any such theor...

Abstract not found

We address the problem of sentence alignment for monolingual corpora, a phenomenon distinct from alignment in parallel corpora. Aligning large comparable corpora automatically would provide a valuable resource for learning of text-totext rewriting rules. We incorporate context into the search for an optimal alignment in two complementary ways: learning rules for matching paragraphs using topic structure and further refining the matching through local alignment to find good sentence pairs. Evaluation shows that our alignment method outperforms state-of-the-art systems developed for the same task.

This report presents a summary of the architectural characteristics of some of the Natural Language Generation (NLG) systems that serve as the main building blocks of the RAGS

This paper addresses this issue in the context of the expression of procedural relations between actions in instructional text. It employs the following four step approach to achieve this goal: (1) Collect a corpus of the relevant text type; (2) Perform a detailed linguistic study of a portion of this corpus, called the training set, and reserving the remainder as a testing set; (3) Implement the results of this study in a text generation system; (4) Compare the output of the system with the text found in the entire corpus. This has resulted in the construction of IMAGENE, an instructional text generation system which embodies a model of the forms of expression consistently used by instructional text writers over a broad range of instruction types. The details of IMAGENE's