<?xml version="1.0" standalone="yes"?>
<Paper uid="W05-0835">
  <Title>A Recursive Statistical Translation Model[?]</Title>
  <Section position="3" start_page="199" end_page="199" type="intro">
    <SectionTitle>
2 Previous works
</SectionTitle>
    <Paragraph position="0"> The initial formulation of the proposed model, including the training procedures, was presented in (Vilar Torres, 1998), along with preliminary experiments in a small translation task which provided encouraging results.</Paragraph>
    <Paragraph position="1"> This model shares some similarities with the stochastic inversion transduction grammars (SITG) presented by Wu in (Wu, 1997). The main point in common is the type of possible alignments considered in both models. Some of the properties of these alignments are studied in (Zens and Ney, 2003). However, the parametrizations of SITGs and the MAR are completely different. The generative process of SITGs produces simultaneously the input and output sentences and the parameters of the model refer to the rules of the nonterminals. This provides a symmetry to both input and output sentences. In contrast, our model clearly distinguishes the input and output sentences and the parameters are based on observable properties of the strings (their lengths and the words composing them). On the other hand, the MAR idea of splitting the sentences until a simple structure is found, also appears in the Divisive Clustering approach presented in (Deng et al., 2004). Again, the main difference lies in the probabilistic modeling of the alignments.</Paragraph>
    <Paragraph position="2"> In Divisive Clustering a uniform distribution on the alignments is assumed while MAR uses a explicit parametrization.</Paragraph>
  </Section>
class="xml-element"></Paper>