<?xml version="1.0" standalone="yes"?>
<Paper uid="C96-2136">
  <Title>Context-Based Spelling Correction for Japanese OCR</Title>
  <Section position="7" start_page="808" end_page="809" type="evalu">
    <SectionTitle>
6 Experiments
</SectionTitle>
    <Paragraph position="0"/>
    <Section position="1" start_page="808" end_page="809" type="sub_section">
      <SectionTitle>
6.1 Language Data and OCR Simulator
</SectionTitle>
      <Paragraph position="0"> We used tile NI'R Dialogue Database (Ehara et el., 1990) to train and test tile spelling correction method. It is a corpus of approximately 800,000 words whose word segmentation anti part ok' speech tagging were laboriously performed by hmu\[. In this experiment, we used one lburth of tile ATR, Corpus, a portion of tile keyboard dialogues in the conference registration domain. 'l'able 1 shows the nmnber of sentences, words, and characters for training anti test data. The test data is not included in the training data. That is, open data were tested in the experiment.</Paragraph>
      <Paragraph position="1">  For the spelling correction experiment, we used an OC, R simulator because it is very difficult to obtain a large amount of test data with arbitrary recognition accuracies. The OCR, simulator takes an input string anti generates a character matrix using a conflmion matrix for Japanese handwriting OCI{,, developed in our laboratory. The parameters of the OCR sinmlator are tile recognition accuracy of the lirst candidate (lirst candklate correct rate), anti tile percentage of tile correct the.r null acters included in tile character matrix (correct candidate included rate).</Paragraph>
      <Paragraph position="2"> In general, the accuracy of current Japanese handwriting OCR is around 90%. It is lower than that of printed characters (around 98%) due to the wide variability in handwriting. When the input comes from FAX, it degrades another 10% to 15%, because tile resolution of most FAX machines is 200dpi, while that of scanners is 400dpi. There\['ore, we made \[bur test sets of' character matrices whose first candidate correct rates and correct candidate included rates were (70%, 90%), (80%, 95%), (90%, 98%), and (95%, 98%), respectively.</Paragraph>
      <Paragraph position="3"> The average numt&gt;er of candidates ibr a character w~s 8.9 in these character matrices 4</Paragraph>
    </Section>
    <Section position="2" start_page="809" end_page="809" type="sub_section">
      <SectionTitle>
6.2 Character Recognition Accuracy
</SectionTitle>
      <Paragraph position="0"> First, we compared the proposed word-based spelling corrector using the POS trigram model (POSe) with tile conventional character I)msed spelling eorreetor using tile character trigram model (Char3). Table 2 shows tile character recognition accuracies after error correction \['or various b~seline OCR accuracies. We also changed the condition of the approximate word match. In Tat)le 2, Matehl, Match2, and Match3 represent that tilt approximate mM;ch fbr substrings whose lengths were more than or equal to one, two, and three characters, respectively.</Paragraph>
      <Paragraph position="1"> In generM, tile approximate match for short words improves character recognition accuracy by about one percent. When the lirst candidate correct rate is low (70% and 80%), tile word based corrector significantly outperIbrnL~ tile character-based corrector. This is because, by approximate word matching, tile word-based corrector can correct words even if the correct, characters are not present in the matrix. When the first candidate correct rate is high (90% and 95%), the word-I&gt;~sed corrector still outperl`orms tile character based eorrector, although the ditDrenee is small.</Paragraph>
      <Paragraph position="2"> This is because most correct characters are al ready included in the ma.trix.</Paragraph>
      <Paragraph position="3">  that the corre.ct candidate included r~ttc increases a.s the tirst candi(hm~ correct rate incrc~Lscs, a.nd that NOllle correct characters ~re l|ev(:r \[)resellt ill tile Illg-trix ewm if the first candidate correct ,:~Lt(~ is high.</Paragraph>
    </Section>
    <Section position="3" start_page="809" end_page="809" type="sub_section">
      <SectionTitle>
6.3 Word Segmentation and Word
Correction Accuracy
</SectionTitle>
      <Paragraph position="0"> First, we deline the performance mea,sures of J apanese word segmentation and word correction.</Paragraph>
      <Paragraph position="1"> We will think of' tile output of tile spelling eorrector ~ a set of 2-tuples, word segmentation and orthography. We then compare tile tuples con tained in the system's output to tile tuptes contained in the standard analysis. For tile N-best candidate, we will make the union of tile tuples contained in each candidate, in other words, we will make a word lattice from N-best candidates, and compare them to tile tuples in the standard.</Paragraph>
      <Paragraph position="2"> For comparison, we count tile number of tuples in tile standard (Std), the number of tuples in the system output (Sys), and tile number of matching tuples (M). We' then calculate recall (M/Std) and precision (M/Sys) as accuracy measures.</Paragraph>
      <Paragraph position="3"> We define two degrees of equality among tuples for counting the number of matching tuples. For word segmentation accuracy, two tuples are equal if they have tile same word segmentation regard less of orthography. For word correction accuracy, two tuples are equal if they have the same word segmentation and orthography.</Paragraph>
      <Paragraph position="4"> Table 5 shows the words segmentation accuracy and word correction accuracy. The word segmen ration accuracy of tile spelling eorrector is signitieantly high, even if the input is very noisy.</Paragraph>
      <Paragraph position="5"> For example, when the accuracy of the baseline OCI{. is 80%, since tile a.verage numlmr of char acters and words in the test sentences are 20.1 and 11.3, there are 4.0 (=20.1'(1-0.80)) chm'actee errors in the sentence, in average. Ilowever, 94.5% word segmentation recall means that there are only 0.62 (=11.3'(1-0.945)) word segmenta tions that are not found in the first candidate.</Paragraph>
      <Paragraph position="6"> Moreover, we t&gt;el the word correction accuracy in Table 3 is satisfactory \['or an interactive spelling corrector. For example, when the accuracy of the b~seline OCI{ is 90%, there are 2.0 (=20.1&amp;quot;(1 0.90)) cha.racter errors in the test sentence, llow ever, 92.8% reca.ll for the first candidate and 95.6% recall for tile top 5 candidates means that there are only 0.81 (11.3&amp;quot;0-0.928)) words that are not found in the lirst candidate, and if you exa.mine the top 5 candidates, this wdue is reduced to 0.50 (~1.3'(1-0.9S@). That is, about half of the er rors in the lirst candidate are corrected by simply selecting tile alternatives in the word lattice.</Paragraph>
    </Section>
  </Section>
class="xml-element"></Paper>