<?xml version="1.0" standalone="yes"?>
<Paper uid="W03-0308">
  <Title>TREQ-AL: A word alignment system with limited language resources</Title>
  <Section position="6" start_page="1" end_page="1" type="evalu">
    <SectionTitle>
5 Evaluation
</SectionTitle>
    <Paragraph position="0"> The results of the evaluation of TREQ-AL performance are shown in the Table 1. In our submission file the sentence no. 221 was left out by (our) mistake. We used the official evaluation program to re-evaluate our submission with the omitted sentence included and the precision improved with 0,09%, recall with 0,45%, F-measure and AER with 0,33%.). The figures in the first and second columns of the Table 1 are those considered by the official evaluation. The last column contains the evaluation of the result that was our main target. Since TREQ-AL produces only &amp;quot;sure&amp;quot; links, AER (alignment error rate - see the Shared Task web-page for further details) reduces to 1 - F-measure.</Paragraph>
    <Paragraph position="1"> TREQ-AL uses no external bilingual-resources. A machine-readable bilingual dictionary would certainly improve the overall performance. The present version of the system (which is far from being finalized) seems to work pretty well on the non-null assignments and this is not surprising, because these links are supposed to be relevant for a translation dictionary extraction system and this was the very reason we developed TREQ.</Paragraph>
    <Paragraph position="2"> Moreover if we consider only the content words (main categories: noun, verbs, adjectives and general adverbs), which are the most relevant with respect to our immediate goals (multilingual wordnets interlinking and word sense disambiguation), we think TREQ-AL performs reasonably well and is worth further improving it.</Paragraph>
  </Section>
class="xml-element"></Paper>