<?xml version="1.0" standalone="yes"?>
<Paper uid="W01-0704">
  <Title>Semantic Pattern Learning Through Maximum Entropy-based WSD techniquea0</Title>
  <Section position="7" start_page="0" end_page="0" type="concl">
    <SectionTitle>
6 Conclusions and outstanding work
</SectionTitle>
    <Paragraph position="0"> In this paper we have presented a semantic pattern learning system driven by a WSD method based on Maximum Entropy models. These semantic patterns have been applied to the anaphora resolution through the construction of ontological patterns. The adding of this pattern learning improve, as it can be seen, the anaphora resolution process. We have pointed out the main advantages of this approach comparing it with other.</Paragraph>
    <Paragraph position="1"> The WSD method is based on conditional Maximum Entropy probability models. It is a supervised learning method that uses a semantically annotated corpus for training. ME models are used in order to estimate functions that performs a sense classification of nouns, verbs and adjectives. The learning phase has been made with simple features with no deep linguistic knowledge. Preliminary results indicate that the accuracy of the model is comparable to other learning methods.</Paragraph>
    <Paragraph position="2"> The main problem in the addition of this kind of knowledge is the lack of appropriate resources to deal with these tasks. In our research work we are trying to apply these techniques both in English and Spanish. The WSD method have been mainly developed in English, but one of our main goals is the design of a complete anaphora resolution system for Spanish. In this way, the main problem is the short available resources regarding to semantically tagged corpora in Spanish (unlike in English). This lack affects the correct development of tasks belonging to the research line shown in this paper, such us the pattern learning and the anaphora resolution. Nevertheless, this shortage opens the door to new research lines that join English resources and multilingual techniques for the generation of patterns in other languages from the learned English patterns.</Paragraph>
  </Section>
class="xml-element"></Paper>