<?xml version="1.0" standalone="yes"?>
<Paper uid="H05-1100">
  <Title>Proceedings of Human Language Technology Conference and Conference on Empirical Methods in Natural Language Processing (HLT/EMNLP), pages 795-802, Vancouver, October 2005. c(c)2005 Association for Computational Linguistics Morphology and Reranking for the Statistical Parsing of Spanish</Title>
  <Section position="7" start_page="801" end_page="801" type="concl">
    <SectionTitle>
6 Conclusions and Future Work
</SectionTitle>
    <Paragraph position="0"> We have developed a statistical parsing model for Spanish that performs at 85.1% F1 constituency accuracy. We find that an approach that explicitly represents some of the particular features of Spanish (i.e., its morphology) does indeed help in parsing. Moreover, this approach is compatible with the reranking approach, which uses general features that were first developed for use in an English parser. In fact, our best parsing model combines both the language-specific morphological features and the non-specific reranking features. The morphological features are local, being restricted to dependencies between words in the parse tree; the reranking features are more global, relying on larger portions of parse structures. Thus, we see our final model as combining the strengths of two complementary approaches.</Paragraph>
    <Paragraph position="1"> We are curious to know the extent to which a close analysis of the dependency errors made by the baseline parser can be corrected by the development of features tailored to addressing these problems.</Paragraph>
    <Paragraph position="2"> Some preliminary investigation of this suggests that we see much higher gains when using generic features than these more specific ones, but we leave a thorough investigation of this to future work. Another avenue for future investigation is to try using a more sophisticated baseline model such as Collins' Model 2, which incorporates both subcategorization and complement/adjunct information. Finally, we would like to use the Spanish parser in an application such as machine translation.</Paragraph>
  </Section>
class="xml-element"></Paper>