<?xml version="1.0" standalone="yes"?>
<Paper uid="P92-1029">
  <Title>Association-based Natural Language Processing with Neural Networks</Title>
  <Section position="9" start_page="228" end_page="229" type="evalu">
    <SectionTitle>
5 Evaluation
</SectionTitle>
    <Paragraph position="0"> To evaluate tile method, we tested the implemented sytem by doing kana-kanji conversion for real documents. The training data and tested data were taken from four types of documents: business letters, personal letters, news articles, and technical articles. The amount of training data and tested data was over 100,000 phrases and 10,000 phrases respectively, for each type of document. The measure for accuracy of conversion was a reduction ratio(RR) of the homonym choice operations of a user. For comparison, we also evaluated the reduction ratio(RR ~) of the kana-kanji conversion with a conventional context holding mechanism.</Paragraph>
    <Paragraph position="2"> A : number of clmice operations required when an untrained kana-kanji converter was used.</Paragraph>
    <Paragraph position="3"> B : number of choice operations required when a NN-trained kana-kanji converter was used.</Paragraph>
    <Paragraph position="4"> C : nunlber of choice operations required when a kana-kanji converter with a conventional context holding mechanism was used. Tile result is shown in Table 1. The advantages of our method is clear for each type  business letters 41.8 32.6 personal letters 20.7 12.7 news articles 23.4 12.2 technical articles 45.6 40.7 of documents. Especially, it is notable that the advantages in business letter field is prominent, because more than 80% of word processor users write business letters.</Paragraph>
  </Section>
class="xml-element"></Paper>