<?xml version="1.0" standalone="yes"?>
<Paper uid="C04-1099">
  <Title>Query Translation by Text Categorization</Title>
  <Section position="6" start_page="0" end_page="0" type="concl">
    <SectionTitle>
5 Conclusion and future work
</SectionTitle>
    <Paragraph position="0"> We have presented a cross language information retrieval engine, which capitalizes on the availability of multilingual controlled vocabulary to translate user requests. The system relies on a text categorizer, which maps queries into a set of predeflned concepts. The automatic text categorizer is tuned to perform a keyword assignment task before being used to translate French queries into English MeSH terms. For OHSUMED queries, optimal precision is obtained when selecting three MeSH terms, but results are improved when the system is merged with a commercial machine translation system, what suggest that text categorization can be opportunely combined with other query translation approaches. As future investigation, we plan to take into account the retrieval status value obtained by each of the ranked MeSH terms instead of simply setting a threshold on the ranking of the terms.</Paragraph>
    <Paragraph position="1"> 9They report on surprisingly better results (CLIR ration = 71%) for Spanish queries and suggest that French is more di-cult to translate than Spanish !</Paragraph>
  </Section>
class="xml-element"></Paper>