<?xml version="1.0" standalone="yes"?>
<Paper uid="W00-0507">
  <Title>TransType: a Computer--Aided Translation Typing System</Title>
  <Section position="6" start_page="49" end_page="49" type="evalu">
    <SectionTitle>
4 Evaluation
</SectionTitle>
    <Paragraph position="0"> We have conducted a theoretical evaluation of TransType on a word completion task, which assumes that a translator carefully observes each completion proposed by the system, and accepts it as soon as it is correct. Under these optimistic conditions, we have shown that TransType allows for the production of a translation typing less than a third of its characters. In order to better grasp the usefulness of TRANSTYPE, we also performed a more practical evaluation by asking ten translators to use the prototype for about one hour to translate isolated sentences. We first asked them to translate without any help from TRANSTYPE and then we compared their typing speed with TRANSTYPE suggestions turned on. Overall, translators liked the concept and found it very useful; they all liked the suggestions although it seemed to induce a literal style of translation. We also asked them if they thought that TRANSTYPE improved their typing speed and the majority of them said so; unfortunately the figures showed that none of them did so ... The typing rates are nevertheless quite good, given that the users were new to this environment and this style of looking at suggestions while translating. But interestingly this practical ew~luation confirmed our theoretical evaluation that atranslation can be produced with TRANSTYPE by typing less than 40% of the characters of a translation. Results of this evaluation and comparisons with our theoretical figures are further described in (Foster et al., 2000).</Paragraph>
    <Paragraph position="1"> This experiment made us realize that this concept of real-time suggestions depends very much on the usability of the prototype; we had first developed a much simpler editor but its limitations were such that the translators found it unusable. So we are convinced that the user-interface aspects of this prototype should be thoroughly studied. But the TRANSTYPE approach would be much more useful if it was combined with other text editing tasks related to translation: for example TRANSTYPE could format the translation in the same way as the source text, this would be especially useful for titles and tables; it would also be possible to localize automatically specific entities such as dates, numbers and amounts of money. It would also be possible to check that some translations given by the user are correct with respect with some normative usage of words or terminological coherence; these facilities are already part of TRANSCHECK, another computer aided translation tool prototype developed in our laboratory (Jutras, 2000).</Paragraph>
  </Section>
class="xml-element"></Paper>