<?xml version="1.0" standalone="yes"?>
<Paper uid="W04-3103">
  <Title>The Language of Bioscience: Facts, Speculations, and Statements in Between</Title>
  <Section position="7" start_page="0" end_page="0" type="concl">
    <SectionTitle>
5 Conclusion and future work
</SectionTitle>
    <Paragraph position="0"> The work presented here is preliminary but promising. It seems that the notion of speculative sentence can be characterized enabling manual annotation. However, we did not manage to characterize the distinction between high and low speculation. In addition, it seems likely that automated systems will be able to achieve useful accuracy. Finally, abstracts seem to include a fair amount of speculative information. null Future work concerning manual annotation would include revising the guidelines, throwing out the High vs. Low distinction, annotating more data, annotating sub-sentential units, annotating the focus of the speculation (e.g., a gene), and annotating full text articles. We are also ignorant of work in linguistics that almost certainly exists and may be informative. We have started this process by considering (Hyland, 1998) and (Harris et al., 1989).</Paragraph>
    <Paragraph position="1"> Future work concerning automatic annotation includes expanding the substring system with more substrings and perhaps more complicated regular expressions, expanding the feature set of the SVM, trying out other classification methods such as decision trees.</Paragraph>
    <Paragraph position="2"> Finally, we plan on building some of the applications mentioned: a speculation search engine, transcription factor interaction tables with a speculation/definite column, and knowledge discovery test sets.</Paragraph>
  </Section>
class="xml-element"></Paper>