<?xml version="1.0" standalone="yes"?>
<Paper uid="C00-2122">
  <Title>Learning to Select a Good Translation</Title>
  <Section position="2" start_page="0" end_page="843" type="intro">
    <SectionTitle>
1 Introduction
</SectionTitle>
    <Paragraph position="0"> Verbmobil (Wahlster, 2000) is a speech to speech machine translation system, aimed at handling a wide range of spontaneous spee('h phenomena within the restricted domain of travel t)lanning and at)pointment sche(hfling dialogues. For the language 1)airs English-German an(1 Gerinan-English, tbur different translation methods are applied in parallel, thus increasing the system's robustness and versatility. Since, exa(:tly one translation should t)(; t)roduce(1 for each int)ut utterance, a selection 1)rocedure is necessary. In order to benelit more from this diversity, the alternative translations are furthermore, combined within the t)omldaries of single utterances, so as to form \]low cornpound translations. Combining translations t'ronl different sources within a multi-thread MT system has already proved beneficial in the past (Frederking and Nire, nlmrg, 1994). Our i)resent work diflhrs fl:om the work reported in there ill several ways (at)art from the trivial fact that we use 'four heads' rather than three). Firstly, we attempt to investigate a systematic solution to |;11(; incolnparability of the various confidence values. Secondly, as we deal with speech to speech rather than text to text translation, dit'ti;renl; segmentations for each given input string are allowed, lnaking the segment combination process signitic~mtly ulore COml)licated.</Paragraph>
    <Section position="1" start_page="0" end_page="0" type="sub_section">
      <SectionTitle>
1.1 Incomparability
</SectionTitle>
      <Paragraph position="0"> Eat;h translation lnodule calculates a confidence value tbr each translation that it; produces.</Paragraph>
      <Paragraph position="1"> However, since the w~rious translation methods are flmdalnentally difl'erent from one another, the resulting contidenee values cannot t)e dire(:tly (:omt)ared across modules. Whereas we do assmne a gener~fl (:orrest)ondence between confidence wflues and translation quality within each one of the mo(hfles, there is no guaranty whatsoever that a high value delivered t)y a certain module wouM indeed signify a l)etter translation when comt)ared with another value, even a much lower one, which was delivered l)y another module. An additional step needs to 1)e taken in order to make the confidence wflues comparable with one another.</Paragraph>
    </Section>
    <Section position="2" start_page="0" end_page="843" type="sub_section">
      <SectionTitle>
1.2 Working Hypotheses
</SectionTitle>
      <Paragraph position="0"> It should be noted that one of our working hypotheses, namely, that coniidence values do generally reflect translation quality, also (:omt)ensates to ~t certain extent tbr tile lack of a wide range theory of translation, according to which translations of difli;rent sorts could be mlanimously ewduated. The task of evaluating  translation quality is non-trivial also for human annotators, since the applicable criteria are diverse, and at the absence of a comprehensive translation theory, very often lead to contradicting conclusions. This difficulty is partially dealt with in section 4.1 below, but tbr practical reasons we tend to accept the need to rely on human judgment, partially theory assisted and partially intuitive, as inevitable. Another presupposition that underlies the current work is that the desirable rescaling Call be well approximated by means of linear polynomials. This assumption allows us to remain within the relatively friendly realm of linear equations (albeit inconsistent), and reflects two basic guiding principles: firstly, that tile rescaling is motivated by pragnlatical needs, rather than by descriptive aspirations, and secondly, that it should not contradict the presupposed correlation between confidence and quality within each module, which implies that the rescaling functions should be monotonous.</Paragraph>
    </Section>
  </Section>
class="xml-element"></Paper>