DARPA Switchboard
		    Telephone Conversation Corpus
				(swb1)

			      Release 2
			     August, 1997


			Disc 1 of 23, Phase 1

The Switchboard-1 Telephone Speech Corpus was originally collected by
Texas Instruments in 1990-1, under DARPA sponsorship.  The first
release of the corpus was published by NIST and distributed by the
LDC in 1992-3.  Since that release, a number of corrections have been
made to the data files as presented on the original CD-ROM set, and
all copies of the first pressing have been distributed.

In this new release, assembled and published by the LDC, all known
errors affecting the original publication of speech files have been
corrected.  In addition, modifications have been made to the contents
of the NIST Sphere headers of all speech files, to identify each file
as being part of the new release, and to make the usage of the
"sample_count" header field consistent with standard Sphere usage.
(In particular, the "sample_count" field should reflect the number of
samples on each channel in the file.  In the initial release, this
field was improperly set to be the total number of samples in both
channels of the file; this has been corrected in the new release.)

Each speech file consists of a 1024-byte ASCII-formatted Sphere
header, followed by 2-channel interleaved mu-law sample data.  The
mu-law samples represent the actual digital data transmission from the
telephone service provider (MCI), as captured separately for each side
of the telephone conversation by an InterVoice RobotOperator
voice-response system.  Full documentation for the data collection is
provided together with the transcripts of the conversations, which are
obtained from the LDC in combination with the CD-ROMs of speech data
(or separately).

As indicated in the volume-id strings of the CD-ROMs, the collection
was organized into two "phases"; calls collected in Phase 1 are on the
first 13 discs of the set, and the Phase 2 calls are on the remaining
10 discs.  The difference between the two phases involved the average
length of conversations (the Phase 2 calls are typically shorter), and
characteristics of the InterVoice collection hardware, which in turn
lead to different post-processing for the Phase 1 calls.

In particular, there was a "feature" in the original InterVoice system
software that caused a limited number of samples to be dropped from the
digital capture under certain conditions in signal quality.  While
this feature was not noticeable to participants while it was taking
place, it resulted in a variable loss of alignment between the two
channels of the call in the resulting digital recording.  Once this
discrepancy between channels was noticed in the speech data, further
data collection was suspended until the InterVoice software was
patched, and a re-alignment process was engineered at NIST to repair
the affected speech files.

Hence, the Phase 1 portion consists of files that have been
manipulated to assure proper time alignment of the two channels, while
the Phase 2 portion contains files that were never affected by
mis-alignment.  The alignment operations performed on the Phase 1
files are logged by way of "*.rec" (reconciliation) files that
accompany the transcripts for those conversations.

In the root directory of each Switchboard-1 Speech DVD-ROM, you will
find:

	docs/readme.txt		(this file)
	docs/swb1_all.dvd.tbl	list of all files modified for the DVD set
	docs/file.tbl	file table with checksums and modification timestamps
	data/			directory containing speech files

The speech files are named according to the following pattern:

	sw0NNNN.sph

where the five-digit string "0NNNN" represents the conversation-id;
this string is used to identify all files associated with the given
call (i.e. transcripts and reconciliation files, as well as speech
data), and also to identify the calls in the associated data base
tables that provide information about the calls and participants.
(The initial zero of the five-digit conversation-id serves to
distinguish the files of Switchboard-1 from those of more recent
Switchboard collections that have been created by the LDC.)