<?xml version="1.0" standalone="yes"?>
<Paper uid="W98-1427">
  <Title>GENERATION AS A SOLUTION TO ITS OWN PROBLEM</Title>
  <Section position="3" start_page="0" end_page="257" type="intro">
    <SectionTitle>
1 Introduction
</SectionTitle>
    <Paragraph position="0"> It Is generally agreed that the technology of natural language generation has evolved to a stage where it can feasibly be expected to be found in ,real world', applied systems. Indeed, within theiast year there has been a specialist tutorial (Dale and Reiter, 1997) and a journal article (Reiter and Dale, 1997) aimed at guiding interested parties on how to build such systems; a textbook on this subject is also about to appear (Reiter and Dale, forthcoming). A problem that remains outstanding, however, is that of the input to NLG applications: where should we get it from and what should it look like (McDonald, 1993)? A popular school of thought on this issue is echoed in the following quotation:  * GOSSIP takes data from an audit trail of an operating system and produces a report (for a security officer) on user activity over the period (Carcagno and Iordanskaja, 1989).</Paragraph>
    <Paragraph position="1"> * FOG takes data from a time series of weather depiction charts and produces a bilingual (French and English) weather report for the period (Goldberg et al, 1994).</Paragraph>
    <Paragraph position="2"> * LFS takes statistical data from labour force surveys and from this produces a report on employment statistics over the given period (Iordanskaja et al, 1992).</Paragraph>
    <Paragraph position="3"> * MODELEXPLAINER takes data from graphical object oriented data models and from this generates a textual description of the model (Lavoie et al, 1996).</Paragraph>
    <Paragraph position="4"> * POSTGRAPHE takes tabular data (of the sort found in a typical spreadsheet) and generates a report integrating both graphics and text (Fasciano and Lapalme, 1996).</Paragraph>
    <Paragraph position="5"> * PLANDOC takes the data from a simulation log file and from this produces a report of the explored simulation options (McKeown et al, 1994).</Paragraph>
    <Paragraph position="6"> * ALETHGEN takes data from a customer database and produces a customised letter (in French)  (Coch, 1996).</Paragraph>
    <Paragraph position="7"> * A system developed by Johanna Moore and her colleagues at the University of Pittsburg takes the data from SAGE, a graphics presentation system (Roth et al, 1994), and produces an accompanying natural language caption (Mittal et al, in press).</Paragraph>
    <Paragraph position="8"> However, it doesn't make the only sense. Some applications require the user to interact rather more closely with the semantic model that drives the generation process, and there is a small, but growing, number of systems that are able to provide this kind of interaction. They achieve this through a common solution: interfaces that allow the user to engage in symbolic authoring of the generated text.</Paragraph>
    <Paragraph position="9"> * EXCLASS is an intelligent support tool for personnel officers writing (bilingual English and French) job descriptions. The user builds the job description by composing and editing conceptual representations; these representations are trees of concepts from a structured conceptual dictionary. Concepts are presented to the user through diagrammatic trees with natural language labels (Caldwell and Korelsky, 1994).</Paragraph>
    <Paragraph position="10"> * DRAFTER-I is an authoring tool to support technical authors and software developers in writing (bilingual English and French) software manuals. The user directly builds the domain model (semantic knowledge base) describing the procedures for using a selected software application. As it is being constructed, the model is presented to the user through diagrams and fragments of text (Paris et al, 1995).</Paragraph>
    <Paragraph position="11"> * GIST is an authoring tool to support forms designers. It generates (multilingual English, German, Italian) forms in the domain of social administration. The user's interaction with GIST is very similar to that with DRAFTER-I. (Power et al, 1995).</Paragraph>
    <Paragraph position="12"> * A tool to support inventors in the authoring of patent claims allows the user to build a semantic model of the invented apparatus by selecting (via multiple-choice menu options) the apparatus parts, their functions and relations to each other (Sheremetyeva and Nirenburg, 1996).</Paragraph>
    <Paragraph position="13"> These systems all comprise a natural language generator coupled to an interface that supports the manual creation of the generator's input (i.e., the authoring of the symbolic (conceptual) content of the output document).</Paragraph>
    <Paragraph position="14"> In the remainder of this paper we describe a new type of symbolic authoring, WYSIWYM (What You See Is What You Meant) editing. A unique feature of WYSIWYM is that it uses the text generator not only to produce the output document but also to contribute to the editing of its own input specification.</Paragraph>
  </Section>
class="xml-element"></Paper>