<?xml version="1.0" standalone="yes"?>
<Paper uid="E06-1007">
  <Title>Automatic Detection of Nonreferential It in Spoken Multi-Party Dialog</Title>
  <Section position="7" start_page="54" end_page="55" type="concl">
    <SectionTitle>
5 Conclusion and Future Work
</SectionTitle>
    <Paragraph position="0"> This paper presented a machine learning system for the automatic detection of nonreferential it in spoken dialog. Given the fact that our feature extraction methods are only very shallow, the results we obtained are satisfying. On the one hand, the good results that we obtained when utilizing information about interruption points (P:80.0% / R:60.9% / F:69.2%) show the feasibility of detecting nonreferential it in spoken multi-party dialog. To our knowledge, this task has not been tackled before. On the other hand, the still fairly good results obtained by only using automatically determined features (P:71.9% / R:55.1% / F:62.4%) show that a practically usable filtering component for nonreferential it can be created even with rather simple means.</Paragraph>
    <Paragraph position="1"> All experiments yielded classifiers that are conservativeinthesensethattheirprecisionisconsid- null erably higher than their recall. This makes them particularly well-suited as filter components.</Paragraph>
    <Paragraph position="2"> For the coreference resolution system that this  work is part of, only the fully automatic variant is an option. Therefore, future work must try to improve its recall without harming its precision (too much). Onewaytodothatcouldbetoimprovethe  recognition(i.e.correctPOStagging)ofgrammaticalfunctionwords(inparticularcomplementizers null like that) which have been shown to be important indicators for constructions with nonreferential it.</Paragraph>
    <Paragraph position="3"> Other points of future work include the refinement  ofthesyntacticpatternfeaturesandthelexicalfeatures. E.g., the values (i.e. mostly nouns, verbs, and adjectives) of the lexical features, which have been almost entirely ignored by both classifiers, could be generalized bymapping them to common WordNet superclasses.</Paragraph>
  </Section>
class="xml-element"></Paper>