<?xml version="1.0" standalone="yes"?>
<Paper uid="J93-3002">
  <Title>A Computational Theory of Goal-Directed Style in Syntax</Title>
  <Section position="7" start_page="493" end_page="495" type="concl">
    <SectionTitle>
6. Summary and Conclusion
</SectionTitle>
    <Paragraph position="0"> In this final section, we will review the contributions of the research described in this paper and conclude with a discussion of some of the new directions that the work is taking.</Paragraph>
    <Section position="1" start_page="493" end_page="494" type="sub_section">
      <SectionTitle>
6.1 Contributions of the Research
</SectionTitle>
      <Paragraph position="0"> The problem of style presented advantages as a focus for new research. The codification of stylistic knowledge had been a virtually unexplored problem even within the general research area of computational linguistics. With very few exceptions, previous work had been unambitious (for example, counting word frequencies, or advocating basic rules of composition).</Paragraph>
      <Paragraph position="1"> Our aim was to create a formal representation of goal-directed, nonliterary stylistics and, moreover, to do so in a manner applicable to different languages. The solution we proposed was the codification of stylistic knowledge in the form of a stylistic grammar. The construction of a stylistic grammar constitutes a theoretical advance over previous work in stylistics, for researchers had not attempted to produce a formal treatment of style, but had relied simply on unstructured normative or descriptive 'rules.' The work we have done toward a grammar of style has brought together ideas from stylistic theory and knowledge representation and applied them to a hitherto unformalized body of knowledge.</Paragraph>
      <Paragraph position="2"> As a result of building English and French stylistic grammars, we were able to give more-formal definitions of stylistic goals. Previously, our understanding had been either purely subjective or based on established but unformalized usage. Now we have a grammar that correlates stylistic goals with specific patterns of abstract properties of text. Our contribution to more-formal definitions of stylistic goals was demonstrated for both syntactic and semantic style, as Ryan (1989, 1992) adapted our vocabulary and methodology to construct a semantic stylistic grammar that correlated the focus structure of paragraphs with the abstract elements, and the abstract elements with specific stylistic goals. In so doing, he augmented the definitions of stylistic goals.</Paragraph>
      <Paragraph position="3"> The English and French syntactic stylistic grammars were implemented in STYLIS- null Chrysanne DiMarco and Graeme Hirst Goal-Directed Style in Syntax TIQUE, a stylistic parser that produced detailed goal-directed stylistic analyses of sentences typical of sophisticated magazine writing.</Paragraph>
    </Section>
    <Section position="2" start_page="494" end_page="494" type="sub_section">
      <SectionTitle>
6.2 Applications of the Research
</SectionTitle>
      <Paragraph position="0"> stylistic grammars and parsers. DiMarco and Hirst (1990) describes the application of our ideas to machine translation. Given the current French and English stylistic grammars, the next step in building a system that could preserve style in translation was to define a mapping between these grammars. Mah (1991, 1992) adapted and extended our work to add a practical, computational treatment of French-English comparative stylistics to the theory. This work will eventually make possible machine translation systems that would be able to preserve or modify style in translation.</Paragraph>
      <Paragraph position="1"> Makuta-Giluk (1991; Makuta-Giluk and DiMarco 1993) developed a computational theory of rhetoric that builds upon our theory to deal with the codification of higher-level pragmatic effects of language, such as formality, persuasion, and sincerity. Her work is also applicable to the preservation of stylistic effects in translation.</Paragraph>
      <Paragraph position="2">  with stylistic and pragmatic aspects of language is important not only in natural language understanding, but in generation as well. Our stylistic grammar provides a formal representation of stylistic knowledge that was previously lacking in generation systems that attempted to deal with pragmatic issues. BenHassine (1992) adapted our knowledge representation for stylistics in order to incorporate stylistic constraints into the Penman language generation system (Penman 1988). Green (1992a, 1992b) extended and refined our theory of style by adapting work from functional grammar.</Paragraph>
      <Paragraph position="3"> Hoyt (1993) and Green have implemented the new theory in an integrated stylistic analyzer and generator, respectively. DiMarco, Hirst, and Stede (1993) are looking at lexical choice in natural language generation, including considerations of lexical style. 6.2.3 Second-Language Teaching. What has been learned from developing STYLISTIQUE has also been applied to machine-aided language instruction. Existing languageteaching systems focus almost exclusively on the basics of composition. An instructional version of STYLISTIQUE could systematically develop a student's understanding of the more advanced aspects of language composition. Payette (1990; Payette and Hirst 1992) developed an instructional system that, applying some of our ideas, analyzes input sentences for basic normative style and clarity, and offers feedback to the student.</Paragraph>
    </Section>
    <Section position="3" start_page="494" end_page="495" type="sub_section">
      <SectionTitle>
6.3 Conclusion
</SectionTitle>
      <Paragraph position="0"> Stylistic and pragmatic aspects, though necessary in complete understanding of language, have been neglected in computational linguistics research. These problems had been too vague and ill-defined to be dealt with by computational systems. However, in this work, we have developed a novel, formal representation of stylistic knowledge that makes the problem of stylistic analysis more amenable to computational solution.</Paragraph>
      <Paragraph position="1"> It is hoped that this research will lead to a system sophisticated enough to deal with a range of stylistic problems. Long-term applications include the development of a stylistic post-editor for use in a machine translation system. In addition, the continuing enhancements of the stylistic analyzer should contribute to a better understanding of the role style plays in language generation and teaching. The ongoing development of a formal framework for the representation of knowledge about stylistics should provide a partial computer model of how people produce style in language.</Paragraph>
      <Paragraph position="2">  Computational Linguistics Volume 19, Number 3</Paragraph>
    </Section>
  </Section>
class="xml-element"></Paper>