<?xml version="1.0" standalone="yes"?>
<Paper uid="W04-3230">
  <Title>Applying Conditional Random Fields to Japanese Morphological Analysis</Title>
  <Section position="2" start_page="0" end_page="0" type="intro">
    <SectionTitle>
1 Introduction
</SectionTitle>
    <Paragraph position="0"> Conditional random fields (CRFs) (Lafferty et al., 2001) applied to sequential labeling problems are conditional models, trained to discriminate the correct sequence from all other candidate sequences without making independence assumption for features. They are considered to be the state-of-the-art framework to date. Empirical successes with CRFs have been reported recently in part-of-speech tagging (Lafferty et al., 2001), shallow parsing (Sha and Pereira, 2003), named entity recognition (Mc-Callum and Li, 2003), Chinese word segmentation (Peng et al., 2004), and Information Extraction (Pinto et al., 2003; Peng and McCallum, 2004).</Paragraph>
    <Paragraph position="1"> Previous applications with CRFs assumed that observation sequence (e.g. word) boundaries are fixed, and the main focus was to predict label At present, NTT Communication Science Laboratories, 2-4, Hikaridai, Seika-cho, Soraku, Kyoto, 619-0237 Japan taku@cslab.kecl.ntt.co.jp sequence (e.g. part-of-speech). However, word boundaries are not clear in non-segmented languages. One has to identify word segmentation as well as to predict part-of-speech in morphological analysis of non-segmented languages. In this paper, we show how CRFs can be applied to situations where word boundary ambiguity exists.</Paragraph>
    <Paragraph position="2"> CRFs offer a solution to the problems in Japanese morphological analysis with hidden Markov models (HMMs) (e.g., (Asahara and Matsumoto, 2000)) or with maximum entropy Markov models (MEMMs) (e.g., (Uchimoto et al., 2001)). First, as HMMs are generative, it is hard to employ overlapping features stemmed from hierarchical tagsets and non-independent features of the inputs such as surrounding words, word suffixes and character types. These features have usually been ignored in HMMs, despite their effectiveness in unknown word guessing.</Paragraph>
    <Paragraph position="3"> Second, as mentioned in the literature, MEMMs could evade neither from label bias (Lafferty et al., 2001) nor from length bias (a bias occurring because of word boundary ambiguity). Easy sequences with low entropy are likely to be selected during decoding in MEMMs. The consequence is serious especially in Japanese morphological analysis due to hierarchical tagsets as well as word boundary ambiguity. The key advantage of CRFs is their flexibility to include a variety of features while avoiding these bias.</Paragraph>
    <Paragraph position="4"> In what follows, we describe our motivations of applying CRFs to Japanese morphological analysis (Section 2). Then, CRFs and their parameter estimation are provided (Section 3). Finally, we discuss experimental results (Section 4) and give conclusions with possible future directions (Section 5).</Paragraph>
  </Section>
class="xml-element"></Paper>