<?xml version="1.0" standalone="yes"?>
<Paper uid="A94-1015">
  <Title>A Successful Case of Computer Aided Translation</Title>
  <Section position="8" start_page="92" end_page="93" type="concl">
    <SectionTitle>
6 Conclusions
</SectionTitle>
    <Paragraph position="0"> For anyone knowing well the fields of Natural Language Processing and Machine Translation, the case I have presented may come as a surprise -- and surprised was I with the results obtained with so simple a method. It can be argued that the pair Portuguese-English makes the translation easy, because, for instance, there is no need for translating most of the gender agreements appearing in Portuguese; and that a book in Mathematics is written in a very restricted kind of language with a relatively small number of different constructions. This is obviously true although some remarks should be made, if one is not trying to have a 100%-correct translation:  1. gender agreements lend themselves to simple treatment in most cases; this is to say that most of the time there will be no need for complex analyses to arrive at their correct translation, 2. the pair Portuguese-English poses some difficult  problems concerning word order, and this would be seldom the case, for instance, with pairs of Romance languages, 3. the need for translating scientific texts in Mathematics, Physics, Chemistry, Medicine and other fields in which formalized and restricted subsets of natural languages are used is probably big enough to make translation tools as those I have developed very interesting, 4. to the best of my knowledge, this is the very first case of a book being successfully translated mostly by computer.</Paragraph>
    <Paragraph position="1"> The main conclusion to draw seems to be that there are particular translation problems that can be solved or partially solved using simple and efficient methods in a computer, much effort being saved this way. Nevertheless, even if one is to take the results described so far as pertaining more to ingenuity than to research (an analysis with which I fully agree, the time spent in writing down and improving the programs being quite negligible), there is an interesting set of questions put by them which I think should be brought forward to those working in this kind of problems. These questions are:  what kind of syntactic analysis (even superficial) is needed to improve these tools, and how to integrate it? the same for semantic analysis, for which language pairs would these tools be in/appropriate, and why? how far can a rewrite rule dictionary as the one described be re-used?</Paragraph>
  </Section>
class="xml-element"></Paper>