<?xml version="1.0" standalone="yes"?>
<Paper uid="W01-0803">
  <Title>Document Structuring a la SDRT</Title>
  <Section position="2" start_page="0" end_page="0" type="intro">
    <SectionTitle>
1 Introduction
</SectionTitle>
    <Paragraph position="0"> Using the terms of (Reiter and Dale, 2000), we consider that the Document Planner architecture is pipelined: first the content determination task does its work, and then the document structuring task takes the result and build a document plan.</Paragraph>
    <Paragraph position="1"> Following the work of (Roussarie, 2000), we adopt SDRT (Asher, 1993; Asher and Lascarides, 1998), which was designed first for text understanding, for the document structuring task1.</Paragraph>
    <Paragraph position="2"> The input to the document structuring component is a set of factual data encoded into a logical form, as in (1).</Paragraph>
    <Paragraph position="4"> This level of representation is supposed to be  and document structuring are pipelined here, while they are interleaved in his work.</Paragraph>
    <Paragraph position="5"> includes a conceptual (language independent) relation, i.e. cause, between the eventsa1a4a3 anda1a9a8 . A document plan is a SDRS. Our goal is to produce a wide range of paraphrases from the same factual data. For example, from the logical form in (1), we want to produce at least all the texts in (2). These texts have different communicative structures and so correspond to different communicative goals. However, these issues will not be addressed here.</Paragraph>
    <Paragraph position="6"> (2) a. Fred left. Therefore, Mary burst into a fit of tears.</Paragraph>
    <Paragraph position="7"> b. Mary burst into a fit of tears because Fred left.</Paragraph>
    <Paragraph position="8"> c. Fred left. His leaving brought Mary into a fit of tears.</Paragraph>
    <Paragraph position="9"> d. Mary burst into a fit of tears. This is due to Fred's leaving.</Paragraph>
    <Paragraph position="10"> To produce paraphrases, we start by producing several document plans (i.e. SDRSs) from the same factual data. The SDRS underlying (2a) is in (3a) in which the discourse relation Result between a42 a3 and a42 a8 expresses the predicate</Paragraph>
    <Paragraph position="12"> volved instead of Resulta15a42 a3a5a42 a8 a19 . The SDRS underlying (2c) is in (3b). It includes the discourse relation Commentary2 defined in (Asher, 1993). To ensure the cohesion of texts, we add the following constraint to his definition: Commentarya15a42 a3a27a5a42 a8a27a19 requires that one element in a42 a8 is coreferent with one element in a42 a3 , as it is the case in (3b) with a1a27a46a47a29a48a1 a3 . In (3b), the causal relation has been reified as the discourse referent f (see section 5). This discourse referent is ex- null When provided as input to a &amp;quot;tactical component&amp;quot; (microplanner and surface realizer), a given SDRS leads to zero, one or several texts. It leads to nothing when there is a lexical (or syntactic) gap in the target language. For example, if there is no verbal predicate semantically equivalent to be due to in the target language, the SDRS underlying (2d) leads to nothing. Similarly, if a SDRS includes a discourse relation which cannot be realized in the target language (e.g. volitional-Result proposed in (Mann and Thompson, 1987) cannot be linguistically realized in French (Danlos, 2001)), it leads to nothing4. A given SDRS leads to several texts when there are several lexicalizations for at least one predicate.</Paragraph>
    <Paragraph position="13"> Thanks to the use of SDRT, we are able to give a formal background to the following assump3In the generation community, causative verbal predicates such as bring into or provoke are considered as elementary ones, although it should not be so. For example, Elixir provokes an allergic reaction is not analyzed and so is simply represented as (allergic-reaction (Elixir)) in (Bouayad-Agha et al., 2000). Whereas, it should get a representation translating x's taking Elixir causes x's having an allergic reaction with a causal relation between two events.</Paragraph>
    <Paragraph position="14"> 4We adopt the position that there exists a set of discourse relations which are likely to be language independent.</Paragraph>
    <Paragraph position="15"> tions generally used in bottom-up documentstructuring approaches: a78 &amp;quot;The content determination mechanism has produced a set of messages which are required to be included in the final document plan&amp;quot; (Reiter and Dale, 2000, p. 114). In formal terms, it translates as follows: a SDRS a42 built from a logical form LF is such that the logical form derived from a42 is logically equivalent to LF. For example, the logical forms derived from the SDRSs in (3a) and (3b) are equivalent to that in (1) modulo axioms which will be presented in section 4.</Paragraph>
    <Paragraph position="16"> a78 &amp;quot;The NLG system has a means of determining what discourse relation (if any) can be used to link two particular messages or component document plans&amp;quot; (Reiter and Dale, 2000, p. 114). Our formal approach is based on reversing the SDRT conditions to establish discourse relations. As an illustration, in SDRT for text understanding, there is the Axiom in (4) for Narration. This axiom states that if Narration holds between two SDRSs</Paragraph>
    <Paragraph position="18"> happens before the main event ofa42</Paragraph>
    <Paragraph position="20"> For text generation, this axiom is reversed in the rule in (5) which is domain and language independent. (5) is taken from (Roussarie, 2000, p. 154).</Paragraph>
    <Paragraph position="21"> (5) a78 If a82 and a82a51a83 are DRS the main eventualities of which are not states, a78 and if the main event of</Paragraph>
    <Paragraph position="23"> This paper is organized as follows. Section 2 gives a crash course in SDRT. Section 3 compares our approach to document structuring to those based on RST. Section 4 explains the axioms needed to lay down the logical equivalence of SDRSs such that (3a) and (3b). Section 5 explains the process for building SDRSs. Section 6 sketches how to generate a text from a SDRS. Section 7 illustrates the document structuring strategy on examples.</Paragraph>
  </Section>
class="xml-element"></Paper>