<?xml version="1.0" standalone="yes"?>
<Paper uid="C04-1121">
  <Title>Sentiment classification on customer feedback data: noisy data, large feature vectors, and the role of linguistic analysis</Title>
  <Section position="9" start_page="71" end_page="71" type="concl">
    <SectionTitle>
5 Conclusion
</SectionTitle>
    <Paragraph position="0"> We have shown that in the very noisy domain of customer feedback, it is nevertheless possible to perform sentiment classification. This can be achieved by using large initial feature vectors combined with feature reduction based on log  An adjectival semantic node modified by a verbal proposition and a pronominal subject. This is in fact the representation for a copular construction of the form &amp;quot;pronoun be adjective to verb...&amp;quot; as in &amp;quot;I am happy to report...&amp;quot; likelihood ratio. A second, more surprising result is that the use of abstract linguistic analysis features consistently contributes to the classification accuracy in sentiment classification. While results like this have been reported in the area of style classification (Baayen et al. 1996, Gamon 2004), they are noteworthy in a domain where stylistic markers have not been considered in the past, indicating the need for more research into the stylistic correlations of affect in text.</Paragraph>
  </Section>
class="xml-element"></Paper>