<?xml version="1.0" standalone="yes"?>
<Paper uid="P04-2011">
  <Title>Beyond N in N-gram Tagging</Title>
  <Section position="2" start_page="0" end_page="0" type="intro">
    <SectionTitle>
1 Introduction
</SectionTitle>
    <Paragraph position="0"> The Hidden Markov Model (HMM) used for part-of-speech (POS) tagging is usually a second-order model, using tag trigrams, implementing the idea that a limited number of preceding tags provide a considerable amount of information on the identity of the current tag. This approach leads to good results. For example, the TnT trigram HMM tagger achieves state-of-the-art tagging accuracies on English and German (Brants, 2000). In general, however, as the model does not consider global context, mistakes are made that concern long-distance syntactic relations.</Paragraph>
  </Section>
class="xml-element"></Paper>