Lexicon

• Lexical Entries and Baseforms
  may have multiple pronunciations & correspond to pauses
• HTK format
• consistent with acoustic & language models
• morphological analysis by ChaSen
• frequent words of Mainichi newspaper articles of 4 years (91-94)

Lexical Coverage

vocabulary size coverage

5000 88.3%

20000 96.4%

24000 97.0%

53000 99.0%

60000 99.2%

101000 99.7%

154000 99.9%