<?xml version="1.0" standalone="yes"?>
<Paper uid="P06-2034">
  <Title>Sydney, July 2006. c(c)2006 Association for Computational Linguistics Discriminative Reranking for Semantic Parsing</Title>
  <Section position="8" start_page="268" end_page="268" type="concl">
    <SectionTitle>
5 Conclusion
</SectionTitle>
    <Paragraph position="0"> We have applied discriminative reranking to semantic parsing, where reranking features are developed from features for reranking syntactic parses based on the coupling of syntax and semantics. The best reranking model significantly improves F-measure on a Robocup coaching task (CLANG) from 82.3% to 85.1%, while it fails to improve the performance on a geography database query task (GEOQUERY).</Paragraph>
    <Paragraph position="1"> Future work includes further investigation of the reasons behind the different utility of reranking for the CLANG and GEOQUERY tasks. We also plan to explore other types of reranking features, such as the features used in semantic role labeling (SRL) (Gildea and Jurafsky, 2002; Carreras and M`arquez, 2005), like the path between a target predicate and its argument, and kernel methods (Collins, 2002b). Experimenting with other effective reranking algorithms, such as SVMs (Joachims, 2002) and MaxEnt (Charniak and Johnson, 2005), is also a direction of our future research.</Paragraph>
  </Section>
class="xml-element"></Paper>