OGI ALPHADIGITS

• Volunteers from USENET given list of six-word alphadigit strings.

• 78,000 total utterances (75 hours) from 3031 speakers.

• 8-bit u-law samples collected at 8 kHz from a T1 line.

	Male	Female	Child	Unknown	Total
No. of Speakers	1419	1533	30	1	2983
No. of Utterances	35680	38585	795	29	75089
No. of Clean Utterances	25284	25700	477	29	51490

• Very similar line conditions to SWITCHBOARD data, but the simpler task and henceforth better transcriptions allow for more accurate phone alignments and frame-state identification.