<?xml version="1.0" standalone="yes"?>
<Paper uid="H01-1041">
  <Title>OTHER LANGUAGES SEMANTIC FRAMES (COMMON COALITION LANGUAGE) SEMANTIC FRAMES (COMMON COALITION LANGUAGE) UNDERSTANDING UNDERSTANDING UNDERSTANDING UNDERSTANDING GENERATION GENERATION GENERATION GENERATION</Title>
  <Section position="6" start_page="7" end_page="7" type="concl">
    <SectionTitle>
4. SUMMARY AND ONGOING WORK
</SectionTitle>
    <Paragraph position="0"> We have described the key features of the CCLINC interlingua-based Korean-to-English translation system which is capable of translating a large quantity of Korean newspaper articles on missiles and chemical biological warfare in real time. Translation quality evaluations on the training and test data indicate that the current system produces translation sufficient for content understanding of a document in the training domains. The key research issues identified from the evaluations include (i) parsing complex sentences, (ii) automated acquisition of word sense disambiguation rules from the training corpus, and (iii) development of discourse module to identify the referents of missing arguments. Our solution to the key technical challenges crucially draws upon the utilization of annotated corpora: For complex sentence parsing, we acquire both rules and rule production probabilities from syntactically annotated corpus. For automated word sense disambiguation, we utilize a sense-tagged corpus to identify various senses of a word, and obtain probabilities for word senses in various contexts. For discourse understanding, we are developing an algorithm for our 2-way speech translation work, [12], and plan to expand the module for document translations.</Paragraph>
  </Section>
class="xml-element"></Paper>