• OGI Alphadigits Corpus
  • A telephone database collected over digital phone lines

  • Strings averaged six words: "E B A 1 Q 2"

  • Most highly confusable pair is "S" and "F" due to loss of high frequencies at telephone bandwidth

• Training Data
  • As a proof of concept we chose to experiment with a single syllable, '_eh_f' (the word "f")

  • 395 utterances chosen from the official training set

  • Includes 459 instances of syllable _eh_f