<?xml version="1.0" standalone="yes"?>
<Paper uid="P84-1077">
  <Title>Semantic Rule Based Text Generation</Title>
  <Section position="7" start_page="379" end_page="379" type="evalu">
    <SectionTitle>
6. Summary and Future Work
</SectionTitle>
    <Paragraph position="0"> The current prototype of the Kafka generator is running and generating text for both paraphrases and responses for the XCALIBUR system. This system demonstrates the feasibility of the semantic transformational method of text generation. The transformational rule formalism allows much simpler specification of diverse syntactic and semantic computations than the hard-coded lisp used in the previous version of the XCALIBUR generator.</Paragraph>
    <Paragraph position="1"> Current work on Kafka is focused on three goals: first, more grammar rules are necessary to increase coverage.</Paragraph>
    <Paragraph position="2"> Now that the basic method has been validated, the grammar will be extended to cover the entire XCALIBUR domain. The second goal is making Kafka more efficient. Most of the system's time is spent trying matches that are doomed to fail. The discrimination network now being added to the system will avoid these pointless matches, providing the speed required in an interactive system like XCALIBUR. The third goal is formalizing the remaining ad hoc phases of the generator. Four of seven major components now use transformations; two of these seem amenable to the transformational method. The case ordering can easily be done by transformations. Converting the semantic processing done by the relat module will be more difficult, since the rule interpreter will have to be expanded to allow multiple inputs.</Paragraph>
  </Section>
class="xml-element"></Paper>