<?xml version="1.0" standalone="yes"?>
<Paper uid="P06-1134">
  <Title>Sydney, July 2006. c(c)2006 Association for Computational Linguistics Word Sense and Subjectivity</Title>
  <Section position="8" start_page="1070" end_page="1071" type="concl">
    <SectionTitle>
6 Conclusion and Future Work
</SectionTitle>
    <Paragraph position="0"> The questions posed in the introduction concerning the possible interaction between subjectivity and word sense found answers throughout the paper. As it turns out, a correlation can indeed be established between these two semantic properties of language.</Paragraph>
    <Paragraph position="1"> Addressing the first question of whether subjectivity is a property that can be assigned to word senses, we showed that good agreement (k=0.74) can be achieved between human annotators labeling the subjectivity of senses. When uncertain cases are removed, the k value is even higher (0.90). Moreover, the automatic subjectivity scoring mechanism that we devised was able to successfully assign subjectivity labels to senses, significantly outperforming an &amp;quot;informed&amp;quot; baseline associated with the task. While much work remains to be done, this first attempt has proved the feasibility of correctly assigning subjectivity labels to the fine-grained level of word senses.</Paragraph>
    <Paragraph position="2"> The second question was also positively answered: the quality of a word sense disambiguation system can be improved with the addition of subjectivity information. Section 5 provided evidence that automatic subjectivity classification may improve word sense disambiguation performance, but mainly for words with both subjective and objective senses. As we saw, performance may even degrade for words that do not. Tying the pieces of this paper together, once the senses in a dictionary have been assigned subjectivity labels, a word sense disambiguation system could consult them to decide whether it should consider or ignore the subjectivity feature.</Paragraph>
    <Paragraph position="3"> There are several other ways our results could impact future work. Subjectivity labels would be a useful source of information when manually augmenting the lexical knowledge in a dictionary,  e.g., when choosing hypernyms for senses or deciding which senses to eliminate when defining a coarse-grained sense inventory (if there is a subjective sense, at least one should be retained).</Paragraph>
    <Paragraph position="4"> Adding subjectivity labels to WordNet could also support automatic subjectivity analysis. First, the input corpus could be sense tagged and the subjectivity labels of the assigned senses could be exploited by a subjectivity recognition tool. Second, a number of methods for subjectivity or sentiment analysis start with a set of seed words and then search through WordNet to find other subjective words (Kamps and Marx, 2002; Yu and Hatzivassiloglou, 2003; Hu and Liu, 2004; Kim and Hovy, 2004; Esuli and Sebastiani, 2005). However, such searches may veer off course down objective paths. The subjectivity labels assigned to senses could be consulted to keep the search traveling along subjective paths.</Paragraph>
    <Paragraph position="5"> Finally, there could be different strategies for exploiting subjectivity annotations and word sense. While the current setting considered a pipeline approach, where the output of a subjectivity annotation system was fed to the input of a method for semantic disambiguation, future work could also consider the role of word senses as a possible way of improving subjectivity analysis, or simultaneous annotations of subjectivity and word meanings, as done in the past for other language processing problems.</Paragraph>
    <Paragraph position="6"> Acknowledgments We would like to thank Theresa Wilson for annotating senses, and the anonymous reviewers for their helpful comments.</Paragraph>
    <Paragraph position="7"> This work was partially supported by ARDA AQUAINT and by the NSF (award IIS-0208798).</Paragraph>
  </Section>
class="xml-element"></Paper>