<?xml version="1.0" standalone="yes"?>
<Paper uid="W93-0110">
  <Title>Acquiring Predicate-Argument Mapping Information from Multilingual Texts</Title>
  <Section position="7" start_page="113" end_page="114" type="concl">
    <SectionTitle>
4 Conclusion
</SectionTitle>
    <Paragraph position="0"> We have automatically built lexicons with predicate-argument mapping information from English, Spanish and Japanese corpora. These lexicons have been used for several multi-lingual data extraction applications (cf. Aone et ai. \[2\]) and a prototype Japanese-English  machine translation system. The algorithms presented here have minimized our lexical acquisition effort considerably.</Paragraph>
    <Paragraph position="1"> Currently we are investigating ways in which thematic role slots of verb frames and semantic type restrictions on these slots can be derived automatically from corpora (cf. Dagan and Itai \[6\], Hindle and Rooth \[10\], Zernik and Jacobs \[20\]) so that knowledge acquisition at all three levels of predicate-argument mapping can be automated.</Paragraph>
  </Section>
class="xml-element"></Paper>