<?xml version="1.0" standalone="yes"?>
<Paper uid="P06-2116">
  <Title>Sydney, July 2006. c(c)2006 Association for Computational Linguistics A Grammatical Approach to Understanding Textual Tables using Two-Dimensional SCFGs</Title>
  <Section position="8" start_page="911" end_page="911" type="concl">
    <SectionTitle>
6 Conclusion
</SectionTitle>
    <Paragraph position="0"> We have introduced a framework to support a more linguistically-oriented approach to finer interpretation of tables, using two-dimensional stochastic CFGs with Viterbi parsing to find appropriate semantic interpretations of textual tables in terms of different data models. This approach yields a concise model that at the same time facilitates broader coverage than existing models, and is more easily scalable and maintainable. We also introduce a cleaner and richer data model to represent semantic interpretations, and illustrate how it systematically captures a wider range of table types. Without such a data model, the right attribute-value relations caanot be extracted from a table, even if surface elements like &amp;quot;header&amp;quot; and &amp;quot;data&amp;quot; are correctly labeled as previous models attempted to do. Our experiments show that even without other ontological and linguistic knowledge, excellent semantic interpretation accuracy canbeobtainedbyparsingwithatwo-dimensional grammar based on these data models, by using a wide variety of surface features in the terminal symbols. We plan next to extend the model by incorporating ontological and linguistic knowledge for additional disambiguation leverage.</Paragraph>
  </Section>
class="xml-element"></Paper>