<?xml version="1.0" standalone="yes"?>
<Paper uid="W97-0502">
  <Title>Langer S. and Hickey M. in preparation. Using Semantic Lexicons for Intelligent Message Retrieval in a Communication Aid. Submitted to Journal of Natural Language Engineering, special issue on Natural Language Processing for Communication</Title>
  <Section position="6" start_page="13" end_page="13" type="concl">
    <SectionTitle>
5 Conclusions
</SectionTitle>
    <Paragraph position="0"> We have detailed the reasons which lead to the design of a communication aid for non-speakers based on ideas from text retrieval with semantic expansion, and we demonstrated the overall design of the prototype. The main differences between standard information retrieval and text retrieval for an AAC system were presented, namely the size and type of texts retrieved by the system and the necessity to minimize the cognitive load, which leads to the minimal input requirement. In the detailed system description, we have shown that a precise morphological analysis can be achieved -- at least for English -- with relatively low effort, ff we use data from publicly available resources. The morphological module is indispensable to be able to enhance the system with a query expansion algorithm, which is needed to satisfy the minimal input requirement for communication aids.</Paragraph>
    <Paragraph position="1"> Several ideas for improving the text retrieval algorithms and WordKeys and their inclusion in other communication aids are still waiting to be realised.</Paragraph>
    <Paragraph position="2"> One idea is to use a semantic lexicon that is able to learn from the input. This would mean that successful semantic links will get a higher weight than other ones. New links could be added based on knowledge of the user's message selections. Interactions where initial search words do not retrieve relevant messages could be recorded.</Paragraph>
    <Paragraph position="3"> A further aim is the integration of the retrieval module with other AAC systems. WordKeys is not designed to assist all kinds of communication. Integration with other AAC software should be investigated, such as software designed for unique text entry (word prediction systems) and for the rapid use of quick conversational fillers.</Paragraph>
  </Section>
class="xml-element"></Paper>