<?xml version="1.0" standalone="yes"?>
<Paper uid="H05-1045">
  <Title>Proceedings of Human Language Technology Conference and Conference on Empirical Methods in Natural Language Processing (HLT/EMNLP), pages 355-362, Vancouver, October 2005. c(c)2005 Association for Computational Linguistics Identifying Sources of Opinions with Conditional Random Fields and Extraction Patterns</Title>
  <Section position="9" start_page="360" end_page="361" type="concl">
    <SectionTitle>
8 Conclusion
</SectionTitle>
    <Paragraph position="0"> We have described a hybrid approach to the problem of extracting sources of opinions in text. We cast this problem as an information extraction task, using both CRFs and extraction patterns. Our research is the first to identify both direct and indirect sources for all types of opinions, emotions, and sentiments.</Paragraph>
    <Paragraph position="1"> Directions for future work include trying to increase recall by identifying relationships between opinions and sources that cross sentence boundaries, and relationships between multiple opinion expressions by the same source. For example, the fact that a coreferring noun phrase was marked as a source in one sentence could be a useful clue for extracting the source from another sentence. The probability or the strength of an opinion expression may also play a useful role in encouraging or suppressing source extraction.</Paragraph>
  </Section>
class="xml-element"></Paper>