<?xml version="1.0" standalone="yes"?>
<Paper uid="X98-1013">
  <Title>INFORMATION EXTRACTION RESEARCH AND APPLICATIONS: CURRENT PROGRESS AND FUTURE DIRECTIONS</Title>
  <Section position="11" start_page="300" end_page="300" type="concl">
    <SectionTitle>
5 Conclusions and Future Directions
</SectionTitle>
    <Paragraph position="0"> We have summarized SRI's developments in addressing two major obstacles to the broad deployment of IE technology: accuracy and portability. The TIPSTER program has witnessed significant progress in both areas, and has perhaps witnessed even greater progress in our understanding of IE technology.</Paragraph>
    <Paragraph position="1"> We believe that the current state of IE technology suggests two main directions for future work; directions which look to opposite directions of the research-to-applications spectrum. The first direction is to leverage the progress we have made to embed IE technology within applications in which it can be useful. Candidate applications include document retrieval, task-based summarization, task-based machine translation, cross-document and multimedia fusion, and trend analysis. Current progress 2Although it should be noted that every sentence needs to be processed up through the combiner phase if coreference is to work optimally, since referents for referential expressions can occur in otherwise irrelevant sentences. The degree to which ignoring this fact affects performance is an empirical question, which will be studied in future work.</Paragraph>
    <Paragraph position="2">  prepares us well for such investigations, the critical question being whether current levels of accuracy are sufficient for success.</Paragraph>
    <Paragraph position="3"> Our work has also suggested that if we are to achieve revolutionary (rather than merely evolutionary) improvements in the state-of-the-art, we also need to step back and focus on fundamental research. Current approaches are good at identifying the information that natural language &amp;quot;wears on its sleeve&amp;quot;; the remainder will require new and richer techniques. Basic research is necessary to guide the development of such mechanisms, and must be coupled with an investigation into evaluation mechanisms. null</Paragraph>
  </Section>
class="xml-element"></Paper>