<?xml version="1.0" standalone="yes"?>
<Paper uid="C04-1058">
  <Title>Kowloon</Title>
  <Section position="7" start_page="0" end_page="0" type="concl">
    <SectionTitle>
5 Conclusion
</SectionTitle>
    <Paragraph position="0"> We have investigated frequently raised questions about N-fold Templated Piped Correction (NTPC), a generalpurpose, conservative error correcting model, which has been shown to reliably deliver small but consistent gains on the accuracy of even high-performing base models on high-dimensional NLP tasks, with little risk of accidental degradation. Experimental evidence shows that when error-correcting high-accuracy base models, simple models and hypotheses are more beneficial than complex ones, while the more complex and powerful models are surprisingly unreliable or damaging in practice.</Paragraph>
  </Section>
class="xml-element"></Paper>