The 1997 HUB5 Arabic Evaluation was produced by the Linguistic Data Consortium (LDC); catalog number LDC2002S22 and ISBN 1-58563-232-5.
The 1997 HUB5 Non-English evaluation is part of an ongoing series of periodic evaluations conducted by NIST. These evaluations provide an important contribution to the direction of research efforts and the calibration of technical capabilities. They are intended to be of interest to all researchers working on the general problem of conversational speech recognition. To this end the evaluation was designed to be simple, to focus on core speech technology issues, to be fully supported, and to be accessible.
The HUB5 Non-English evaluation, conducted in the fall of 1997, complemented another related evaluation which was conducted in the spring. The spring evaluation focuses on the recognition of conversational speech in English. This evaluation is dedicated to the advancement of speech recognition technology for languages other than English; specifically for Arabic, German, Mandarin, and Spanish. It focuses also on issues related to porting recognition technology to new languages, to system generality, and to language commonalties and universals.
The HUB5 Non-English evaluation focuses on the task of transcribing conversational speech into text. This task is posed in the context of conversational telephone speech in Arabic, German, Mandarin, and Spanish. The evaluation is designed to foster research progress, with the goals of:
- exploring promising new ideas in the recognition of conversational speech
- developing advanced technology incorporating these ideas, and
- measuring the performance of this technology
The task is to transcribe conversational speech. The speech to be transcribed is presented as a set of conversations collected over the telephone. Each conversation is represented as a "4-wire" recording, that is with two distinct sides, one from each end of the telephone circuit. Each side is recorded and stored as a standard telephone codec signal (8 kHz sampling, 8-bit mu-law encoding).
Each conversation is represented as a sequence of "turns," where each turn is the period of time when one speaker is speaking. Each successive turn results from a reversal of speaking and listening roles for the conversation participants. The transcription task is to produce the correct transcription for each of the specified turns. The beginning and ending times of each of these turns will be supplied as side information to the system under test. This turn information will be supplied in NDX format, with one NDX file for all conversations to be transcribed. (Note that the turns are not necessarily a simple sequence of non-overlapping time intervals. They may be overlapping or non-alternating from time to time, because there is no sequencing constraint on conversational interaction.)
Additional documentation is available at the 1997 NIST Evaluation Plan for Recognition of Conversational Speech Over the Telephone website.
This publication contains 20 sphere files encoded in two channel interleaved mulaw with a sampling rate of 8 KHz, for a total of 424,160,000 bytes (405 Mbytes) of sphere data. The sphere headers have been modified from the original Evaluation data by the addition of sample checksums to the CALLHOME data files.
An included documentation table contains information on the speech segments to be processed as follows: ...
There are no updates at this time.
Portions © 1996-2002 Trustees of the University of Pennsylvania.