<?xml version="1.0" standalone="yes"?>
<Paper uid="C04-1045">
  <Title>Improving Word Alignment Quality using Morpho-syntactic Information</Title>
  <Section position="8" start_page="0" end_page="0" type="concl">
    <SectionTitle>
7 Conclusions
</SectionTitle>
    <Paragraph position="0"> In this work we have presented an approach for including morpho-syntactic knowledge into a maximum likelihood training of statistical translation models. As can be seen in Section 5, going beyond full forms during the training by taking into account the interdependencies of the different derivations of the same base form results in the improvements of the alignment  baseline system (name of the model) and the system using hierarchical method (+hier) quality, especially for the small training corpus. We assume that the method can be very effective for cases where only small amount of data is available. We also expect further improvements by performing a special modelling for the rare words.</Paragraph>
    <Paragraph position="1"> We are planning to investigate possibilities of improving the alignment quality for different language pairs using different types of morpho-syntactic information, like for example to use word stems and suffixes for morphologicaly rich languages where some parts of the words have to be aligned to the whole English words (e.g.</Paragraph>
    <Paragraph position="2"> Spanish verbs, Finnish in general, etc.) We are also planning to use the refined alignments for the translation process.</Paragraph>
  </Section>
class="xml-element"></Paper>