<?xml version="1.0" standalone="yes"?>
<Paper uid="N03-2023">
  <Title>Category-Based Pseudowords</Title>
  <Section position="6" start_page="4" end_page="4" type="concl">
    <SectionTitle>
5 Conclusions
</SectionTitle>
    <Paragraph position="0"> We have shown that creating pseudowords based on distributions from lexical category co-occurrence can produce a more accurate lower-bound for WSD systems that use pseudowords than the standard approach. This method allows for the detailed study of a particular sense ambiguity set since many different pseudowords can be generated from one category pair. Additionally, this method provides a better-motivated basis for the grouping of words into pseudowords, since they more realistically model the meaning similarity patterns of real ambiguous words than do randomly paired words.</Paragraph>
    <Paragraph position="1"> Acknowledgements Special thanks to Barbara Rosario for the discussions and valuable suggestions and to Ariel Schwartz for providing the abbreviation extraction code. This work was supported by a gift from Genentech and an ARDA Aquaint contact.</Paragraph>
  </Section>
class="xml-element"></Paper>