<?xml version="1.0" standalone="yes"?>
<Paper uid="W06-2605">
  <Title>Discourse Parsing: Learning FOL Rules based on Rich Verb Semantic Representations to automatically label Rhetorical Relations</Title>
  <Section position="5" start_page="38" end_page="39" type="concl">
    <SectionTitle>
5 Conclusions and Future Work
</SectionTitle>
    <Paragraph position="0"> We have shown that it is possible to learn First Order Logic rules from complex semantic data using an ILP based methodology. These rules can be used to automatically label rhetorical relations.</Paragraph>
    <Paragraph position="1"> Moreover, ourresultsshowthataDiscourseParser that uses only semantic information can perform as well as the state of the art Discourse Parsers based on syntactic and lexical information.</Paragraph>
    <Paragraph position="2"> Future work will involve the use of syntactic informationaswell. Wealsoplantorunamorethorough evaluation on the complete set of relations that we have used in our coding scheme. It is also important that the manual segmentation and annotation of rhetorical relations be subject to inter-annotator agreement. A second human annotator is currently annotating a sample of the annotated corpus. Upon completion, the annotated corpus will be checked for reliability.</Paragraph>
    <Paragraph position="3">  DatasparsenessisawellknownprobleminMachine Learning. Like most paradigms, our learning model is also affected by it. We also plan to explore techniques to deal with this issue.</Paragraph>
    <Paragraph position="4">  Lastly, we have not tackled the problem of discourseparsingathigherlevelsoftheDPTandseg- null mentation in this paper. Our ultimate goal is to build a Discourse Parser that will automatically segment a full text as well as annotate it with rhetorical relations at every level of the DPT using semantic as well as syntactic information. Much work needs to be done but we are excited to see what the aforesaid future work will yield.</Paragraph>
  </Section>
class="xml-element"></Paper>