<?xml version="1.0" standalone="yes"?>
<Paper uid="E85-1036">
  <Title>A RULE-BASED APPROACH TO EVALUATING IMPORTANCE IN DESCRIPTIVE TEXTS</Title>
  <Section position="3" start_page="244" end_page="244" type="intro">
    <SectionTitle>
2. EVALUATING IMPORTANCE
</SectionTitle>
    <Paragraph position="0"> The topic of importance evaluation has been dealt with in recent years, although often only in a quite indirect way, by several authors and in many different contexts. A part of a text can be considered important in relation to other segments of the same text according to several criteria: it embodies knowledge necessary to understand other parts of the text (van Dijk, 1977; Kintsch and van Dijk, 1978); - it is relevant to the topic of discourse (Lehnert 1982 and 1984); it is useful to clarify the relations that make discourse coherent (Hobbs, 1982); it relates to the topic-focus articulation (Haji~ova' and Sgall, 1984); it refers to objects or relations in the subject domain that are judged to be important a-priori (Schank, 1979); - it is unusual, new, or abnormal in the subject domain (Schank, 1979); it generates surprise (van Dijk and Kintsch, 1983); it is relevant to some specific reader's goal or need (Fum, Guida, and Tasso, 1982).</Paragraph>
    <Paragraph position="1"> In practice, if we test these criteria on sample texts, they result sometimes complementary, sometimes partially overlapping, sometimes even conflicting. Moreover, different readers may judge differently the importance of the same text; on some parts a general consensus may be achieved, but the evaluation of other parts may be definitely subjective.</Paragraph>
    <Paragraph position="2"> In fact, &amp;quot;important&amp;quot; means &amp;quot;specially relevant to some goal&amp;quot;, and, whenever the goal with which a text is read changes, the parts of text which are to be considered important vary accordingly. Even if the goal of reading is only seldom considered explicitly by humans, still some goal is always implicitly assumed. Different readers (or the same reader in different moments) may have different goals, and conflicting judgments of importance may be due to the consideration of different goals, rather than to the application of different evaluation procedures.</Paragraph>
    <Paragraph position="3"> The above investigation shows that importance is a really multifaceted concept which escapes a simple, explicit, algorithmic definition. A procedural, knowledge-based approach comprising a set of rules that can assign relative importance values to the different parts of a text and can resolve or explain conflicting evaluations seems more appropriate. Such an approach allows taking into account in a flexible and natural way the variety of knowledge sources and processing activities that are involved in importance evaluation. Moreover, it is expected to be well founded from a cognitive point of view (van Dijk and Kintsch, 1983; Anderson, 1976), as it allows close and transparent modeling of several processes that occur in human mind.</Paragraph>
  </Section>
class="xml-element"></Paper>