<?xml version="1.0" standalone="yes"?>
<Paper uid="H93-1028">
  <Title>THE MURASAKI PROJECT: MULTILINGUAL NATURAL LANGUAGE UNDERSTANDING</Title>
  <Section position="9" start_page="147" end_page="147" type="concl">
    <SectionTitle>
4. CONCLUSION
</SectionTitle>
    <Paragraph position="0"> We have described a multilingual system, Murasaki, focusing on specifics of its language-independent architecture and describing how language-specific data is integrated with general processing modules. While this architecture is currently operating for data extraction from Japanese, Spanish, and English texts, it has been designed to be extended to additional languages in the future. Murasaki also has associated multilingual data acquisition tools and algorithms, which have been used to extend its data modules. In addition, we have developed preliminary multilingual training and evaluation tools for the syntax and discourse modules of Murasaki.</Paragraph>
    <Paragraph position="1"> Planned future enhancements include addition of new data modules (e.g. multilingual &amp;quot;WordNets&amp;quot;), extension of the Spanish and Japanese data sources to new domains, and improved multilingual tools for automatic data acquisition from corpora. We would also like to extend the system to a new, typologically different language such as Arabic in order to further test and refine its language independence.</Paragraph>
  </Section>
class="xml-element"></Paper>