SHACHI - Language Resource Metadata Database

LDC97S44 - Speech data LDC97S66 - Dev and eval LDC97T22 - Transcripts

*Introduction*

The 1996 Broadcast News Speech Corpus contains a total of 104 hours of broadcasts from ABC, CNN and CSPAN television networks and NPR and PRI radio networks with corresponding transcripts. The primary motivation for this collection is to provide training data for the DARPA "HUB4" Project on continuous speech recognition in the broadcast domain.

*Data*

The speech files are available in a 19 disc training data set with one additional disc of development data and an additional disc of evaluation data. The following programs are represented in this corpus:

* ABC Nightline
* ABC World Nightly News
* ABC World News Tonight
* CNN Early Edition
* CNN Early Prime News
* CNN Headline News
* CNN Prime Time News
* CNN The World Today
* CSPAN Washington Journal
* NPR All Things Considered
* NPR Marketplace Transcripts have been made of all recordings in this publication, manually time aligned to the phrasal level, annotated to identify boundaries between news stories, speaker turn boundaries and gender information about the speakers. The released version of the transcripts is in SGML format and there is accompanying documentation and an SGML DTD file, included with the transcription release. The transcripts are available via FTP.

*Updates*

There are no updates at this time.

*Pricing*

The Reduced Licensing Fee for this corpus is US$600.

*Samples*

* audio(MS Wave format).

LDC97S44 - Speech data LDC97S66 - Dev and eval LDC97T22 - Transcripts

*Introduction*

The 1996 Broadcast News Speech Corpus contains a total of 104 hours of broadcasts from ABC, CNN, and CSPAN television networks and NPR and PRI radio networks with corresponding transcripts. The primary motivation for this collection is to provide training data for the DARPA "HUB4" Project on continuous speech recognition in the broadcast domain. The speech files are available in a 19 disc training data set with one additional disc of development data and an additional disc of evaluation data. The following programs are represented in this corpus:

* ABC Nightline
* ABC World Nightly News
* ABC World News Tonight
* CNN Early Edition
* CNN Early Prime News
* CNN Headline News
* CNN Prime Time News
* CNN The World Today
* CSPAN Washington Journal
* NPR All Things Considered
* NPR Marketplace

*Data*

Transcripts have been made of all recordings in this publication, manually time aligned to the phrasal level, annotated to identify boundaries between news stories, speaker turn boundaries and gender information about the speakers. The released version of the transcripts is in SGML format and there is accompanying documentation and an SGML DTD file, included with the transcription release. The transcripts are available via FTP.

*Updates*

There are no updates at this time.

*Samples*

* Text

*Pricing*

The Reduced Licensing Fee for this corpus is US$100.

LDC98S71 - Speech data LDC98T28 - Transcripts

*Introduction*

This release contains a total of 97 hours of recordings from radio and television news broadcasts, gathered between June 1997 and February 1998. It has been prepared to serve as a supplement to the 1996 Broadcast News Speech collection (consisting of over 100 hours of similar recordings). The primary motivation for this collection is to provide additional training data for the DARPA "HUB4" Project on continuous speech recognition in the broadcast domain.

*Data*

Transcripts have been made of all recordings in this publication, manually time aligned to the phrasal level, annotated to identify boundaries between news stories, speaker turn boundaries and gender information about the speakers. The transcription conventions are described in the file "transcrp.doc" -- please note that this file describes the transcription methods by reference to text formatting conventions used internally by the LDC during the transcription process. The released version of the transcripts is in SGML format, comparable to the format that was used in the 1996 Broadcast News Speech transcriptions and there is accompanying documentation and an SGML DTD file, included with the transcription release.

*Updates*

There are no updates at this time.

*Pricing*

The Reduced Licensing Fee for this corpus is US$600.

LDC98S71 - Speech data LDC98T28 - Transcripts

*Introduction*

This publication has been prepared to serve as a supplement to the 1996 Broadcast News Speech collection (consisting of over 100 hours of similar recordings). The primary motivation for this collection is to provide additional training data for the DARPA "HUB4" Project on continuous speech recognition in the broadcast domain.

*Data*

This set of 18 CD-ROMs contains a total of 97 hours of recordings from radio and television news broadcasts, gathered between June 1997 and February 1998.

Transcripts have been made of all recordings in this publication, manually time aligned to the phrasal level, annotated to identify boundaries between news stories, speaker turn boundaries and gender information about the speakers. The transcription conventions are described in the file "transcrp.doc" -- please note that this file describes the transcription methods by reference to text formatting conventions used internally by the LDC during the transcription process. The released version of the transcripts is in SGML format, comparable to the format that was used in the 1996 Broadcast News Speech transcriptions and there is accompanying documentation and an SGML DTD file, included with the transcription release.

*Updates*

There are no updates at this time.

*Pricing*

The Reduced Licensing Fee for this corpus is US$100.

*Introduction*

1997 HUB4 Broadcast News Evaulation Non-English Test Material was developled by the Linguistic Data Consortium. It contains the evaluation test material used in the 1997 DARPA/NIST Continuous Speech Recognition Broadcast News HUB4 Non-English Benchmark Test administered by the NIST Spoken Natural Language Processing Group.

*Data*

The test material is contained in two SPHERE-formatted waveform files. The file h4ne97sp.sph (set1) contains one hour of Spanish broadcast news excerpts from 1997. The file h4ne97ma.sph (set2) contains one hour of Mandarin broadcast news excerpts from 1997. Each file should be separately recognized per the HUB4 Non English Evaluation Specification.

Note: 1997 HUB4 English evaluation material is contained in 1997 HUB4 English Evaluation Speech and Transcripts LDC2002S11.

*Updates*

There are no updates at this time.

*Pricing*

The Reduced Licensing Fee for this corpus is US$150.

*Introduction*

The 1997 HUB4 English Evaluation, Linguistic Data Consortium (LDC) catalog number LDC2002S11 and ISBN 1-58563-227-9, 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 purpose of this evaluation is to foster research on the problem of accurately transcribing broadcast news speech and to measure objectively the state of the art.

Additional documentation is available at the 1997 NIST Evaluation Plan for Broadcast News webpage.

*Data*

This year the entire test set is contained in a single waveform file. The SPHERE-formatted waveforem file h4e_97.sph is located in the h4e_evl directory of this CD-ROM. The waveform file contains 334 Mbytes of sphere data, which represents approximately three hours of concatenated radio and television broadcast news stories. The transcript file contains a rough number of 30,600 total tokens and 4,800 unique tokens.

*Updates*

There are no updates at this time.

*Additional Copyright Information and Disclaimers*

C-SPAN: Washington Journal USC : Radio Marketplace ABC : Prime Time CNN : Morning News CNN : World View NPR : Morning Edition PRI : The World"NPR Programs or excerpts of Programs contained herein are from 1996 broadcasts of "All Things Considered(R)" and "Morning Edition(R)" and are copyrighted works owned by National Public Radio, Inc. The Programs are used with permission of National Public Radio, Inc. Any unauthorized duplication or other use is strictly prohibited."

"RESTRICTED RIGHTS LEGEND: INFORMATION FROM THE USC PROGRAM 'MARKETPLACE' CONTAINED HEREIN IS THE PROPERTY OF USC RADIO AND THE UNIVERSITY OF SOUTHERN CALIFORNIA AND IS PROTECTED BY COPYRIGHT. USE, DUPLICATION OR DISCLOSURE BY YOU IS SUBJECT TO THE RESTRICTIONS SET FORTH IN THE USER AGREEMENT AND ATTACHED TO THE COMPUTER READABLE MEDIA PROVIDED TO YOU BY THE LINGUISTIC DATA CONSORTIUM OF THE UNIVERSITY OF PENNSYLVANIA. COPYRIGHT 1996 UNIVERSITY OF SOUTHERN CALIFORNIA, AND IS DSITRIBUTED TO PUBLIC RADIO STATIONS NATIONWIDE BY PRI-PUBLIC RADIO INTERNATIONAL. MARKETPLACE IS MADE POSSIBLE BY GE, THE CORPORATION FOR PUBLIC RADIO, AND PUBLIC RADIO STATIONS NATIONWIDE."

*Introduction*

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.

*Data*

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:

...

*Updates*

There are no updates at this time.

*Introduction*

The 1997 HUB5 Arabic Transcripts corpus was produced by the Linguistic Data Consortium (LDC), catalog number LDC2002T39 and ISBN 1-58563-245-7.

This publication contains transcripts for twenty CALLHOME Egyptian Arabic telephone conversations. These 20 conversations were used in NIST's 1997 HUB5 Non-English evaluation, and are published as 1997 HUB5 Arabic Evaluation LDC2002S22.

*Data*

There are 40 data files. Each of the 20 calls has transcripts in two formats: .txt and .scr.

The .txt files are transcript files containing the Romanized orthographic forms that were used in the original transcription process. These forms also serve as the head-words in the associated Egyptian Colloquial Arabic Lexicon LDC99L22.

The .scr files are transcript files rendered in Arabic script orthography, using the ISO 8859-6 character set; these files were derived from the .txt files by replacing each word token with its Arabic script counterpart (which is also provided in the CALLHOME Arabic Lexicon). These files have been formatted to avoid problems of bi-directional text: line-feed characters are used to separate ASCII content from Arabic script content in each utterance.

Please follow these links for sample transcripts: txt | scr

*Updates*

There are no updates at this time.

*Introduction*

The 1997 HUB5 German Evaluation was produced by Linguistic Data Consortium (LDC) catalog number LDC2002S24 and ISBN 1-58563-234-1.

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 of that year. 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, and specifically this year 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
* 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 u-law encoding).

Additional documentation is available at the NIST website.

*Data*

This publication contains 20 sphere files encoded in two channel interleaved mulaw with a sampling rate of 8 KHz, for a total of 561,150,160 bytes (535 Mbytes) or nine hours of sphere data.

An included documentation table contains information on the speech segments to be processed as follows:

...

*Updates*

There are no updates at this time.

*Introduction*

The 1997 HUB5 German Transcripts corpus was produced by the Linguistic Data Consortium (LDC), catalog number LDC2003T03 and ISBN 1-58563-247-3.

This publication contains transcripts for 20 CALLHOME German telephone conversations. These twenty conversations were used in NIST's 1997 HUB5 Non-English evaluation, and are published as 1997 HUB5 German Evaluation (LDC2002S24).

*Data*

There are 20 data files in .txt format.

The .txt files are transcript files containing the orthographic forms that were used in the original transcription process. These forms also serve as the head-words in the associated Callhome German Lexicon (LDC97L18).

Please follow this link for a sample transcript.

*Updates*

There are no updates at this time.