<?xml version="1.0" standalone="yes"?>
<Paper uid="W06-2409">
  <Title>Modeling Monolingual and Bilingual Collocation Dictionaries in Description Logics</Title>
  <Section position="6" start_page="69" end_page="69" type="concl">
    <SectionTitle>
5 Conclusion
</SectionTitle>
    <Paragraph position="0"> We have described a model for monolingual and bilingual collocation dictionaries in OWL DL.</Paragraph>
    <Paragraph position="1"> This formalism is well suited for the intended modularization of linguistic resources, be they language- or language-pair- specific (our dictio- null translations and paraphrases, along with results nary models), generalized over one or more languages (our lexicon model), or more abstract, in the sense of a meta-model or an inventory of the descriptive devices shared by the linguistic descriptions of several languages (our model of descriptions, see figure 5 above). This model of descriptions will be larger for related languages (e.g. the indo-european ones), and smaller for typologically very diverse languages; it is however by no means meant to have any interlingual, let alone universal function, but is rather understood in the sense of PARGRAM's shared inventory of descriptive devices9.</Paragraph>
    <Paragraph position="2"> We have modelled so far about 1000 collocations, their components, preferences and relations (also with single words); we intend to considerably enlarge the collocation dictionary, using the possibilities to combine OWL DL models with databases, offered by the Sesame framework. The formalism also supports experiments with credulous inferencing at the level of translation equivalence, e.g. by following not only explicit equivalence relations, but also synonymy relations: in line with the query discussed in section 4.3 above (cf. Figure 7), one could also start from the English express criticism and retrieve the equivalent collocation Kritik &amp;quot;uben as well as its (quasi-)synonyms kritisieren (single word) and in der Luft zerreissen (idiom), which may thus be proposed as equivalent candidates for express criticism.</Paragraph>
    <Paragraph position="3"> More such investigations into the data collec9Cf. http://www2.parc.com/istl/groups/ nltt/pargram/gram.html tion are planned; they may require non-standard access to the dictionary, i.e. access via paths involving other properties and relations than just lemmas and equivalents. The relational nature of the dictionary supports this kind of exploration; we intend to specify and implement a 'linguistfriendly' query overlay to SeRQL and a Graphical User Interface to make such explorations more easy.</Paragraph>
  </Section>
class="xml-element"></Paper>