<?xml version="1.0" standalone="yes"?>
<Paper uid="W02-0801">
  <Title>A Multilingual Approach to Disambiguate Prepositions and Case Suffixes</Title>
  <Section position="8" start_page="2" end_page="2" type="concl">
    <SectionTitle>
6 Conclusion and further work
</SectionTitle>
    <Paragraph position="0"> This paper presents preliminary experiments in the use of translation equivalences to disambiguate prepositions or case suffixes. The core of the method is to find translations of the occurrence of the target preposition or case suffix, and assign the intersection of their set of interpretations. The method is fully automatic, given a table with prepositions and their possible interpretations.</Paragraph>
    <Paragraph position="1"> We have tested this method on the occurrences of the Basque instrumental case -z in the definitions of a Basque dictionary. We have searched the translations in the definitions from 3 Spanish and 3 English dictionaries.</Paragraph>
    <Paragraph position="2"> The results have been that we are able to disambiguate with 94.5% accuracy 2.3% of those occurrences (up to 91). The ambiguity is reduced from 7 readings down to 3.1. We think that these are very good results, especially seeing that there is room for improvement.</Paragraph>
    <Paragraph position="3"> More specifically, we plan to apply surface syntactic analysis to better extract the dependency relations, which is the main source of errors. We would like to study other inventories of preposition interpretations, both in order to have better theoretical foundations as well as to investigate whether coarser grained distinctions would lead to a reduction in the ambiguity.</Paragraph>
    <Paragraph position="4"> In the future, we plan to explore the possibility to feed a Machine Learning algorithm with the automatically disambiguated examples, in order to construct a full-fledged disambiguation algorithm following a bootstrapping approach. On the other hand, we would like to apply the method to the set of all prepositions and case suffixes, and beyond that to all syntactic dependencies. The results will be directly loaded in a Lexical Knowledge Base extracted from the Basque dictionary (Ansa et al., in prep.).</Paragraph>
    <Paragraph position="5"> We also plan to explore whether this method can be applied to free running text, removing the constraint that the translations have to be definitions of the equivalent word.</Paragraph>
    <Paragraph position="6"> Finally, this technique could be coupled with techniques that make use of the semantic types of the words in the context.</Paragraph>
    <Paragraph position="7"> Overall, we found the results are very encouraging given the simple techniques used, and we think that it shows great potential for improvement and interesting avenues for research.</Paragraph>
  </Section>
class="xml-element"></Paper>