<?xml version="1.0" standalone="yes"?>
<Paper uid="W05-1510">
  <Title>Probabilistic models for disambiguation of an HPSG-based chart generator</Title>
  <Section position="7" start_page="99" end_page="101" type="concl">
    <SectionTitle>
6 Conclusion
</SectionTitle>
    <Paragraph position="0"> We presented a chart generator using HPSG and developed log-linear models which we believe was essential to develop a sentence generator. Several techniques developed for parsing also worked in generation. The introduced techniques were an estimation method for log-linear models using a packed forest representation of HPSG trees and iterative beam search. The system was evaluated through application to real-world texts. The experimental results  showed that the generator was able to output a sentence for over 90% of the test data when the data was limited to short sentences. The accuracy was significantly improved by incorporating syntactic features into the log-linear model. As future work we intend to tune the feature set for generation. We also plan to further increase the efficiency of the generator so as to generate longer sentences.</Paragraph>
  </Section>
class="xml-element"></Paper>