<?xml version="1.0" standalone="yes"?>
<Paper uid="P06-1115">
  <Title>Using String-Kernels for Learning Semantic Parsers</Title>
  <Section position="9" start_page="919" end_page="919" type="concl">
    <SectionTitle>
5 Conclusions
</SectionTitle>
    <Paragraph position="0"> We presented a new kernel-based approach to learn semantic parsers. SVM classifiers based on string subsequence kernels are trained for each of the productions in the meaning representation language. These classifiers are then used to compositionally build complete meaning representations of natural language sentences. We evaluated our system on two real-world corpora. The results showed that our system compares favorably to other existing systems and is particularly robust to noise.</Paragraph>
  </Section>
class="xml-element"></Paper>