<?xml version="1.0" standalone="yes"?>
<Paper uid="W03-1702">
  <Title>Class Based Sense Definition Model for Word Sense Tagging and Disambiguation</Title>
  <Section position="6" start_page="0" end_page="0" type="concl">
    <SectionTitle>
6. Conclusion
</SectionTitle>
    <Paragraph position="0"> In this paper, we present the Mutual Assured Resolution of Sense (MARS) Algorithm for assigning relevant senses to word classes in a given sense inventory (i.e. LDOCE or WordNet). We also describe the SWAT Algorithm for automatic sense tagging of a parallel corpus.</Paragraph>
    <Paragraph position="1"> We carried out experiments on an implementation of the MARS and SWAT Algorithms for all the senses in LDOCE and LLOCE. Evaluation on a set of 14 highly ambiguous words showed that very high precision CBSDM and CBSTM can be constructed. High applicability and precision rates were achieved, when applying CBSTM to sense tagging of a Chinese-English parallel corpus.</Paragraph>
    <Paragraph position="2"> A number of interesting future directions present themselves. First, it would be interesting to see how effectively we can broaden the coverage of CBSTM via backing off smoothing. Second, a CBSTM trained directly on a parallel corpus would be more effective in word alignment and sense tagging. The approach of training CBSTM on the L2 glosses in a bilingual MRD may lead to occasional mismatch between MRD translations and in-context translations. Third, there is a lack of research for a more abstractive and modular representation of sense differences and commonality.</Paragraph>
    <Paragraph position="3"> There is potential of developing Sense Definition Model to identify and represent semantic and stylistic differentiation reflected in the MRD glosses pointed out in DiMarco, Hirst and Stede (1993).</Paragraph>
    <Paragraph position="4"> Last but not the least, it would be interesting to apply MARS to both LDOCE E-C and WordNet and project WordNet's sense inventory to a sencond language via CBSDM and a parallel corpus, thus creating a Chinese WordNet and semantic concordance.</Paragraph>
  </Section>
class="xml-element"></Paper>