lr,7-3-H94-1014,ak The <term> models </term> were constructed using a 5K <term> vocabulary </term> and trained using a 76 million word <term> Wall Street Journal text corpus </term> .
hide detail