<?xml version="1.0" standalone="yes"?>
<Paper uid="W91-0207">
  <Title>Representation of Semantic Knowledge with Term Subsumption Languages</Title>
  <Section position="2" start_page="0" end_page="67" type="intro">
    <SectionTitle>
1 Introduction
</SectionTitle>
    <Paragraph position="0"> The lexicon as one of the major linguisitic knowledge sources of a natural language processing (NLP) system contains among others morphological, syntactic and semantic knowledge. The design of an appropriate lexicon involves the following questions: * Which size is required for the lexicon for a given application? * What are the units of the vocabulary? * How should lexical information be represented? * What kinds of techniques should be applied for generating a large lexicon, e.g., could the lexicon be partially generated by extracting information from machine-readable dictionaries? In this paper we address the third problem, more precisely the representation of semantic information.</Paragraph>
    <Paragraph position="1"> Semantic knowledge is crucial for both understanding and generating natural language. One important class of representation models for semantic knowledge are the network models, which have been influenced by early association models. In network models, word meanings can be described by relationships to other word meanings. In contrast, other models for representing semantic knowledge aim at the representation of structural aspects of word meanings so that a particular word meaning is built up by a set of particular semantic features.</Paragraph>
    <Paragraph position="2"> During the last years, a number of knowledge representation (KR) systems have been built using term subsumption languages (TSLs). TSLs are formal languages for defining concepts by reference to superconeepts and by specification of additional features. Because of the similarity of concept descriptions in TSLs and dictionary definitions based on &amp;quot;genus  proximum et differentia specifica', it seems useful to investigate to what extent TSLs can be used for the representation of word meanings.</Paragraph>
    <Paragraph position="3"> In the next section, we discuss some properties of TSLs and their relation to other KR formalisms. In order to get some hints whether TSLs are suitable for the representation of word meanings we examine several dictionary definitions. Next, we give a short introduction to our terminological formalism and outline, how the formalism can be integrated into a lexicon. Finally, we summarize the results and give a short survey of our current work.</Paragraph>
  </Section>
class="xml-element"></Paper>