SRI's Amex Travel Agent Data

Introduction

For purposes of research into dialogue, particularly in the air travel domain used by the ATIS project , SRI audiotaped and transcribed conversations made at SRI International, Menlo Park, CA, between SRI International employees and American Express travel agents. The later transcripts omit portions of the tapes that were not relevant to the ATIS project, noting "[break in transcript]" where this occurs.

Jacqueline Kowtko prepared the original transcripts under the direction of Patti Price. Candy Sidner then edited the transcripts to remove identifying personal information like names and credit card numbers. Ruth Lang has done some additional editing of this kind. Harry Bratt wrote the automatic HTML formatter for these files.

Transcription Conventions

These are the original notes describing the conventions used in transcribing the tapes:

ORTHOGRAPHIC TRANSCRIPTION OF AIRLINE TAPE #1, 5/4/89

Note:   () enclose portions where two or more speakers are talking
	   simultaneously,
	 - means the speaker corrected or switched sentences,
	?? means I'm not sure of the word,
       X/Y means I'm not sure whether X or Y was what was really said,

       Commas used rather inconsistently.
       There's a space between lines whenever the speaker switches.

	[] encloses laugh or sigh,
       ,,, signifies pause,
       xxx means some nonsense sound(s)

Transcriptions

The tapes were transcribed with explicitly noting which utterances belonged to the agent and which did not, as well as noting when a new call started. Because this information was present in the original ascii text file, we were able to automatically parse and format these files in html to make them a bit more readable, and to allow indexing by call. Links to the original ascii versions are below this table. Please note that the date of recording for tapes 17-18 and 20-21 is no longer available and appears as 0 in the table below.

HTML-ized Transcripts
RecordedTape No.Calls
04-24-1989 Tape 1 Call 1 (53 turns) Call 2 (38 turns)
05-04-1989 Tape 3 Call 1 (15 turns) Call 2 (11 turns) Call 3 (36 turns)
05-26-1989 Tape 6 Call 1 (46 turns) Call 2 (18 turns) Call 3 (16 turns) Call 4 (11 turns) Call 5 (14 turns) Call 6 (38 turns)
06-01-1989 Tape 7 Call 1 (11 turns) Call 2 (31 turns) Call 3 (16 turns) Call 4 (17 turns)
06-09-1989 Tape 9ab Call 1 (12 turns) Call 2 (11 turns) Call 3 (28 turns) Call 4 (14 turns)
06-12-1989 Tape 10ab Call 1 (44 turns) Call 2 (17 turns) Call 3 (49 turns)
06-15-1989 Tape 11ab Call 1 (22 turns) Call 2 (14 turns) Call 3 (10 turns)
06-19-1989 Tape 13 Call 1 (46 turns) Call 2 (5 turns) Call 3 (31 turns) Call 4 (26 turns) Call 5 (22 turns)
06-22-1989 Tape 14a Call 1 (23 turns) Call 2 (12 turns) Call 3 (9 turns) Call 4 (9 turns)
06-23-1989 Tape 14b Call 1 (47 turns) Call 2 (26 turns) Call 3 (25 turns)
06-26-1989 Tape 15a Call 1 (20 turns) Call 2 (17 turns) Call 3 (44 turns)
06-30-1989 Tape 15b Call 1 (44 turns) Call 2 (49 turns)
07-06-1989 Tape 16 Call 1 (14 turns) Call 2 (25 turns) Call 3 (19 turns) Call 4 (6 turns) Call 5 (31 turns) Call 6 (25 turns) Call 7 (17 turns) Call 8 (76 turns) Call 9 (59 turns)
00-00-1989 Tape 17 Call 1 (20 turns) Call 2 (112 turns) Call 3 (48 turns) Call 4 (22 turns)
00-00-1989 Tape 18 Call 1 (32 turns) Call 2 (22 turns) Call 3 (24 turns) Call 4 (6 turns) Call 5 (33 turns)
00-00-1989 Tape 20 Call 1 (50 turns) Call 2 (28 turns) Call 3 (15 turns) Call 4 (24 turns)
00-00-1989 Tape 21 Call 1 (47 turns) Call 2 (12 turns)

Here are links to all of the ascii transcriptions corresponding to the html versions in the table above.

Original ASCII-only Transcripts
Tape 1
Tape 3
Tape 6
Tape 7
Tape 9
Tape 10
Tape 11
Tape 13
Tape 14a
Tape 14b
Tape 15a
Tape 15b
Tape 16
Tape 17
Tape 18
Tape 20
Tape 21

Research Based on these Transcripts

Send e-mail to Harry Bratt (username bratt at the host speech.sri.com) if you have a paper using these transcripts that we could list here.