<?xml version="1.0" standalone="yes"?>
<Paper uid="W06-0905">
  <Title>Evaluating Knowledge-based Approaches to the Multilingual Extension of a Temporal Expression Normalizer</Title>
  <Section position="7" start_page="35" end_page="35" type="concl">
    <SectionTitle>
6 Conclusions
</SectionTitle>
    <Paragraph position="0"> In this paper, the problem of automatically extending to new languages a rule-based system for TE recognition and normalization has been addressed.</Paragraph>
    <Paragraph position="1"> Adopting an incremental approach, different porting strategies, for the creation of an Italian system starting from an already available Spanish system, have been evaluated and discussed. Each experiment has been carried out considering the contribution of different knowledge sources for rules translation. Firstly, the contribution given by the output of online translators has been evaluated, showing detection performances in line with a previously developed English extension of the system, but a performance drop in terms of normalization performance. Then, the contribution of knowledge mined from an annotated corpus has been considered. Results show a performance increase in terms of detection and bracketing, but a less coherent behaviour in terms of normalization. Finally, a combined approach has been experimented, resulting in an overall performance increase. System's performance is still far from the results obtained by a state-of-the-art system for Italian but, considering the trade-off between performance and effort required for system's development, results are encouraging.</Paragraph>
  </Section>
class="xml-element"></Paper>