<?xml version="1.0" standalone="yes"?>
<Paper uid="P06-2081">
  <Title>Whose thumb is it anyway? Classifying author personality from weblog text</Title>
  <Section position="9" start_page="632" end_page="633" type="concl">
    <SectionTitle>
7 Conclusion and next steps
</SectionTitle>
    <Paragraph position="0"> This paper has reported the first stages of our investigations into classification of author personality from weblog text. Results are quite promising, and comparable across all four personality traits. It seems that even a small selection of features found to exhibit an empirical relationship with personality traits can be used to generate reasonably accurate classification results. Naturally, there are still many paths to explore. Simple regression analyses are reported in Nowson (2006); however, for classification, a more thorough comparison of different machine learning methodologies is required. A richer set of features besides n-grams should be checked, and we should not ignore the potential effectiveness of unigrams in this task (Pang et al., 2002). A completely new test set can be gathered, so as to further guard against overfitting, and to explore systematically the effects of the amount of training data available for each author. And as just discussed, comparison with human personality classification accuracy is potentially very interesting.</Paragraph>
    <Paragraph position="1"> However, it does seem that we are making progress towards being able to deal with a realistic task: if we spot a thumbs-up review in a weblog, we should be able to check other text in that weblog, and tell whose thumb it is; or more accurately, what kind of person's thumb it is, anyway.</Paragraph>
    <Paragraph position="2"> And that in turn should help tell us how high the thumb is really being held.</Paragraph>
  </Section>
class="xml-element"></Paper>