<?xml version="1.0" standalone="yes"?>
<Paper uid="P06-1028">
  <Title>Training Conditional Random Fields with Multivariate Evaluation Measures</Title>
  <Section position="8" start_page="222" end_page="223" type="relat">
    <SectionTitle>
6 Related Work
</SectionTitle>
    <Paragraph position="0"> Various loss functions have been proposed for designing CRFs (Kakade et al., 2002; Altun et al., 2003). This work also takes the design of the loss functions for CRFs into consideration. However, we proposed a general framework for designing these loss function that included non-linear loss functions, which has not been considered in previous work.</Paragraph>
    <Paragraph position="1"> With Chunking, (Kudo and Matsumoto, 2001) reported the best F-score of 93.91 with the voting of several models trained by Support Vector Machine in the same experimental settings and with the same feature set. MCE-F with the MAP parameter initialization achieved an F-score of 94.03, which surpasses the above result without manual parameter tuning.</Paragraph>
    <Paragraph position="2"> With NER, we cannot make a direct comparison with previous work in the same experimental settings because of the different feature set, as described in Sec. 5.2. However, MCE-F showed the better performance of 85.29 compared with (Mc-Callum and Li, 2003) of 84.04, which used the MAP training of CRFs with a feature selection architecture, yielding similar results to the MAP results described here.</Paragraph>
  </Section>
class="xml-element"></Paper>