<?xml version="1.0" standalone="yes"?>
<Paper uid="W03-0412">
  <Title>PhraseNet: Towards Context Sensitive Lexical Semantics/</Title>
  <Section position="7" start_page="0" end_page="0" type="concl">
    <SectionTitle>
6 Discussion and Further Work
</SectionTitle>
    <Paragraph position="0"> We believe that progress in semantics and in developing lexical resources is a prerequisite to any significant progress in natural language understanding. This work makes a step in this direction by introducing a context-sensitive lexical semantic knowledge base system, PhraseNet. We have argued that while current lexical resources like WordNet are invaluable, we should move towards contextually sensitive resources.</Paragraph>
    <Paragraph position="1"> PhraseNet is designed to fill this gap, and our preliminary experiments with it are promising.</Paragraph>
    <Paragraph position="2"> PhraseNet is an ongoing project and is still in its preliminary stage. There are several key issues that we are currently exploring. First, given that PhraseNet draws part of it power from corpora, we are planning to enlarge the corpus used. We believe that the data size is very important and will add significant robustness to our current results. At the same time, since constructing PhraseNet relies on machine learning techniques, we need to study extensively the effect of tuning these on the reliability of PhraseNet. Second, there are several functionalities and access functions that we are planning to augment PhraseNet with. Among those is the ability of allowing a user to query PhraseNet even without explicitly specifying the role of words in the context. This would reduce the requirement for users and applications using PhraseNet. Finally, current PhraseNet has no lexical information about adjectives and adverbs, which may contain important distributional information about their modified nouns or verbs. We would like to take this information into consideration in the near future.</Paragraph>
  </Section>
class="xml-element"></Paper>