<?xml version="1.0" standalone="yes"?>
<Paper uid="P05-1039">
  <Title>and smoothing</Title>
  <Section position="9" start_page="319" end_page="320" type="concl">
    <SectionTitle>
7 Conclusions
</SectionTitle>
    <Paragraph position="0"> In this paper, we presented the best-performing parser for German, as measured by labelled bracket scores. The high performance was due to three factors: (i) treebank transformations (ii) an integrated model of morphology in the form of a suffix analyzer and (iii) the use of smoothing in an unlexicalized grammar. Moreover, there are possible paths for improvement: lexicalization could be added to the model, as could some of the treebank transformations suggested by Schiehlen (2004). Indeed, the suffix analyzer could well be of value in a lexicalized model.</Paragraph>
    <Paragraph position="1"> While we only presented results on the German NEGRA corpus, there is reason to believe that the techniques we presented here are also important to other languages where lexicalization provides little benefit: smoothing is a broadly-applicable technique, and if difficulties with lexicalization are due to sparse lexical data, then suffix analysis provides a useful way to get more information from lexical elements which were unseen while training.</Paragraph>
    <Paragraph position="2"> In addition to our primary results, we also provided a detailed error analysis which shows that PP attachment and co-ordination are problematic for our parser. Furthermore, while POS tagging is highly accurate, the error analysis also shows it does  have surprisingly large effect on parsing errors. Because of the strong impact of POS tagging on parsing results, we conjecture that increasing POS tagging accuracy may be another fruitful area for future parsing research.</Paragraph>
  </Section>
class="xml-element"></Paper>