<?xml version="1.0" standalone="yes"?>
<Paper uid="C04-1006">
  <Title>Improved Word Alignment Using a Symmetric Lexicon Model</Title>
  <Section position="7" start_page="4" end_page="4" type="relat">
    <SectionTitle>
6 Related Work
</SectionTitle>
    <Paragraph position="0"> The popular IBM models for statistical machine translation are described in (Brown et al., 1993). The HMM-based alignment model was introduced in (Vogel et al., 1996). A good overview of these models is given in (Och and Ney, 2003). In that article Model 6 is introduced as the loglinear interpolation of the other models. Additionally, state-of-the-art results are presented for the Verbmobil task and the Canadian Hansards task for various configurations. Therefore, we chose them as baseline. Compared to our work, these publications kept the training of the two translation directions strictly separate whereas we integrate both directions into one symmetrized training. Additional linguistic knowledge sources such as dependency trees or parse trees were used in (Cherry and Lin, 2003) and (Gildea, 2003). In (Cherry and Lin, 2003)aprobabilitymodelPr(aJ1jfJ1 ;eI1)is used, which is symmetric per definition. Bilingual bracketing methods were used to produce a word alignment in (Wu, 1997). (Melamed, 2000) uses an alignment model that enforces  (Toutanova et al., 2002), extensions to the HMM-based alignment model are presented.</Paragraph>
  </Section>
class="xml-element"></Paper>