<?xml version="1.0" standalone="yes"?>
<Paper uid="E06-2012">
  <Title>Maytag: A multi-staged approach to identifying complex events in textual data</Title>
  <Section position="7" start_page="133" end_page="133" type="concl">
    <SectionTitle>
5 Discussion and future work
</SectionTitle>
    <Paragraph position="0"> The ability to combine a range of analytic processing tools, and the ability to explore their results interactively are the backbone of our approach. In this paper, we've covered the framework of our Maytag prototype, and have looked under its hood at our extraction and classification methods, especially as they apply to financial texts. Much new work is in the offing.</Paragraph>
    <Paragraph position="1"> Many experiments are in progress now to assess performance on other text types (financial news), and to pin down performance on a wider range of events, relations, and structured entities.</Paragraph>
    <Paragraph position="2"> Another question we would like to address is how best to manage the interaction between classification and extraction: a mutual feedback process may well exist here.</Paragraph>
    <Paragraph position="3"> We are also concerned with supporting financial analysis across multiple documents. This has implications in the area of cross-document coreference, and is also leading us to investigate visual ways to define queries that go beyond the paragraph and span many texts over many years.</Paragraph>
    <Paragraph position="4"> Finally, we are hoping to conduct user studies to validate our fundamental assumption. Indeed, this work presupposes that interactive application of multi-purpose classification and extraction techniques can model complex events as well as monolithic extraction tools a la MUC.</Paragraph>
  </Section>
class="xml-element"></Paper>