SHACHI - Language Resource Metadata Database

*Introduction*

This corpus was used as part of the training set for the Second Speech in Noisy Environments Evaluation (SPINE2). SPINE2 provides a continuing forum for assessing the state of the art and practice in speech recognition technology for noisy military environments and for exchanging information on innovative speech recognition technology in the context of fully implemented systems that perform realistic tasks. The evaluation will provide researchers, potential sponsors, and customers with a quantitative means to appreciate the strengths and weaknesses of the technologies. Also, the results reported on will invite customer interest in the potential utility of the technologies. More information on this evaluation is available here.

This work was sponsored in part by National Science Foundation Grant No. IIS-9982201.

*Data*

This publication contains the Speech in Noisy Environments 2 (SPINE2) Clean and Vocoded Training Audio Corpus created for the Department of Defense (DoD) Digital Voice Processing Consortium (DDVPC) by Arcon Corp., and produced by the Linguistic Data Consortium (LDC) as catalog number LDC2001S04 with ISBN 1-58563-206-6. The transcripts for this publication are available as Speech in Noisy Environments (SPINE2) Training Transcripts LDC2001T05 with ISBN 1-58563-207-4. For an example transcript, please click here. These corpora support the 2001 Speech in Noisy Environments evaluation.

The training data comprises two talker pairs (four speakers total) with 32 conversations (sessions) per talker pair (64 conversations total).

The audio for each session is presented in three forms:

* Unprocessed: the signal recorded at the participant's microphone
* Bitstream: the compressed "channel" data produced by the vocoder's analysis stage for transmission from sender to receiver
* Processed: the signal produced by the vocoder's synthesis stage, given the bitstream data as input.
There are a total of 64 clean audio files and 64 vocoded files, one "game" each, for a rough total of seven hours of audio data, 1.6Gb (including the unprocessed, the processed, and the bitstream files), 20,850 total tokens (730 unique tokens).

*Updates*

There are no updates at this time.

*Introduction*

This corpus was used as part of the training set for the Second Speech in Noisy Environments Evaluation (SPINE2). SPINE2 provides a continuing forum for assessing the state of the art and practice in speech recognition technology for noisy military environments and for exchanging information on innovative speech recognition technology in the context of fully implemented systems that perform realistic tasks. The evaluation will provide researchers, potential sponsors, and customers with a quantitative means to appreciate the strengths and weaknesses of the technologies. Also, the results reported on will invite customer interest in the potential utility of the technologies. More information on this evaluation is available here.

This work was sponsored in part by National Science Foundation Grant No. ISS-9982201.

*Data*

This publication contains the Speech in Noisy Environments 2 (SPINE2) Training Transcripts, created for the Department of Defense (DoD) Digital Voice Processing Consortium (DDVPC) by Arcon Corp., and produced by the Linguistic Data Consortium (LDC) as catalog number LDC2001T05 with ISBN 1-58563-207-4. For an example transcript, please click here. The audio for this publication is available as Speech in Noisy Environments (SPINE2) Training Audio LDC2001S04, ISBN 1-58563-206-6. These corpora support the 2001 Speech in Noisy Environments evaluation.

The training data comprises two talker pairs (four speakers total) with 32 conversations (sessions) per talker pair (64 conversations total).

The audio for each session is presented in three forms:

* Unprocessed: the signal recorded at the participant's microphone
* Bitstream: the compressed "channel" data produced by the vocoder's analysis stage for transmission from sender to receiver
* Processed: the signal produced by the vocoder's synthesis stage, given the bitstream data as input.
There are a total of 64 clean audio files and 64 vocoded files, one "game" each, for a rough total of seven hours of audio data, 1.6Gb (including the unprocessed, the processed, and the bitstream files), 20,850 total tokens (730 unique tokens).

*Updates*

There are no updates at this time.

*Introduction*

This corpus was used as the development set for the Second Speech in Noisy Environments Evaluation (SPINE2). SPINE2 provides a continuing forum for assessing the state of the art and practice in speech recognition technology for noisy military environments and for exchanging information on innovative speech recognition technology in the context of fully implemented systems that perform realistic tasks. The evaluation will provide researchers, potential sponsors, and customers with a quantitative means to appreciate the strengths and weaknesses of the technologies. Also, the results reported on will invite customer interest in the potential utility of the technologies. More information on this evaluation is available here.

This work was sponsored in part by National Science Foundation Grant No. IIS-9982201.

*Data*

This publication contains the Speech in Noisy Environments 2 (SPINE2) Clean and Vocoded Development Audio Corpus created for the Department of Defense (DoD) Digital Voice Processing Consortium (DDVPC) by Arcon Corp., and produced by the Linguistic Data Consortium (LDC) as catalog number LDC2001S06 with ISBN 1-58563-208-2. The transcripts for this publication are available as Speech in Noisy Environments (SPINE2) Development Transcripts LDC2001T07 with ISBN 1-58563-209-0. For an example transcript, please click here. These corpora support the 2001 Speech in Noisy Environments evaluation.

The development data comprises two talker pairs (four speakers total) with 16 conversations (sessions) per talker pair (32 conversations total).

The audio for each session is presented in three forms:

* Unprocessed: the signal recorded at the participant's microphone
* Bitstream: the compressed "channel" data produced by the vocoder's analysis stage for transmission from sender to receiver
* Processed: the signal produced by the vocoder's synthesis stage, given the bitstream data as input.
There are a total of 32 clean audio files and 32 vocoded files, one "game" each, for a rough total of three and a half hours (207 minutes) of audio data, 811Mb (including the unprocessed, the processed, and the bitstream files), 9,700 total tokens (600 unique tokens).

*Updates*

There are no updates at this time.

*Introduction*

This corpus was used as the development set for the Second Speech in Noisy Environments Evaluation (SPINE2). SPINE2 provides a continuing forum for assessing the state of the art and practice in speech recognition technology for noisy military environments and for exchanging information on innovative speech recognition technology in the context of fully implemented systems that perform realistic tasks. The evaluation will provide researchers, potential sponsors, and customers with a quantitative means to appreciate the strengths and weaknesses of the technologies, and the results reported on will invite customer interest in the potential utility of the technologies. More information on this evaluation is available here.

This work was sponsored in part by National Science Foundation Grant No. IIS-9982201.

*Data*

This publication contains the Speech in Noisy Environments 2 (SPINE2) Development Transcripts, created for the Department of Defense (DoD) Digital Voice Processing Consortium (DDVPC) by Arcon Corp., and produced by the Linguistic Data Consortium (LDC) as catalog number LDC2001T07 with ISBN 1-58563-209-0. For an example transcript, please click here. The audio for this publication is available as Speech in Noisy Environments (SPINE2) Development Audio LDC2001S06, ISBN 1-58563-208-2. These corpora support the 2001 Speech in Noisy Environments evaluation.

The development data comprises two talker pairs (four speakers total) with 16 conversations (sessions) per talker pair (32 conversations total).

The audio for each session is presented in three forms:

* Unprocessed: the signal recorded at the participant's microphone
* Bitstream: the compressed "channel" data produced by the vocoder's analysis stage for transmission from sender to receiver
* Processed: the signal produced by the vocoder's synthesis stage, given the bitstream data as input.
There are a total of 32 clean audio files and 32 vocoded files, one "game" each, for a rough total of three and a half hours (207 minutes) of audio data, 811Mb (including the unprocessed, the processed, and the bitstream files), 9,700 total tokens (600 unique tokens).

*Updates*

There are no updates at this time.

*Introduction*

This corpus was used as the evaluation set for the Second Speech in Noisy Environments Evaluation (SPINE2). SPINE2 provides a continuing forum for assessing the state of the art and practice in speech recognition technology for noisy military environments and for exchanging information on innovative speech recognition technology in the context of fully implemented systems that perform realistic tasks. The evaluation will provide researchers, potential sponsors, and customers with a quantitative means to appreciate the strengths and weaknesses of the technologies, and the results reported on will invite customer interest in the potential utility of the technologies. More information on this evaluation is available here.

This work was sponsored in part by National Science Foundation Grant No. IIS-9982201.

*Data*

This publication contains the Speech in Noisy Environments 2 (SPINE2) Clean and Vocoded Evaluation Audio Corpus created for the Department of Defense (DoD) Digital Voice Processing Consortium (DDVPC) by Arcon Corp., and produced by the Linguistic Data Consortium (LDC) as catalog number LDC2001S08 with ISBN 1-58563-210-4. The transcripts for this publication are available as Speech in Noisy Environments (SPINE2) Evaluation Transcripts LDC2001T09 with ISBN 1-58563-211-2. For an example transcript, please click here. These corpora support the 2001 Speech in Noisy Environments evaluation.

The evaluation data comprises 16 talker pairs (32 speakers total) with four conversations (sessions) per talker pair (64 conversations total).

The audio for each session is presented in three forms:

* Unprocessed: the signal recorded at the participant's microphone
* Bitstream: the compressed "channel" data produced by the vocoder's analysis stage for transmission from sender to receiver
* Processed: the signal produced by the vocoder's synthesis stage, given the bitstream data as input.
There are a total of 64 clean audio files and 64 vocoded files, one "game" each, for a rough total of seven hours (423 minutes) of audio data, 1.6Gb (including the unprocessed, the processed, and the bitstream files), 23,300 total tokens (930 unique tokens).

*Updates*

There are no updates at this time.

*Introduction*

This corpus was used as the evaluation set for the Second Speech in Noisy Environments Evaluation (SPINE2). SPINE2 provides a continuing forum for assessing the state of the art and practice in speech recognition technology for noisy military environments and for exchanging information on innovative speech recognition technology in the context of fully implemented systems that perform realistic tasks. The evaluation will provide researchers, potential sponsors, and customers with a quantitative means to appreciate the strengths and weaknesses of the technologies, and the results reported on will invite customer interest in the potential utility of the technologies. More information on this evaluation is available here.

This work was sponsored in part by National Science Foundation Grant No. IIS-9982201.

*Data*

This publication contains the Speech in Noisy Environments 2 (SPINE2) Evaluation Transcripts, created for the Department of Defense (DoD) Digital Voice Processing Consortium (DDVPC) by Arcon Corp., and produced by the Linguistic Data Consortium (LDC) as catalog number LDC2001T09 with ISBN 1-58563-211-2. For an example transcript, please click here. The audio for this publication is available as Speech in Noisy Environments (SPINE2) Training Audio LDC2001S08, ISBN 1-58563-210-4. These corpora support the 2001 Speech in Noisy Environments evaluation.

The evaluation data comprises 16 talker pairs (32 speakers total) with four conversations (sessions) per talker pair (64 conversations total).

The audio for each session is presented in three forms:

* Unprocessed: the signal recorded at the participant's microphone
* Bitstream: the compressed "channel" data produced by the vocoder's analysis stage for transmission from sender to receiver
* Processed: the signal produced by the vocoder's synthesis stage, given the bitstream data as input.
There are a total of 64 clean audio files and 64 vocoded files, one "game" each, for a rough total of seven hours (423 minutes) of audio data, 1.6Gb (including the unprocessed, the processed, and the bitstream files), 23,300 total tokens (930 unique tokens).

*Updates*

There are no updates at this time.

*Introduction*

This publication contains the Speech in Noisy Environments 1 (SPINE1) Coded Audio Corpus created for the Department of Defense (DoD) Digital Voice Processing Consortium (DDVPC) by Arcon Corp., and produced by the Linguistic Data Consortium (LDC) as catalog number LDC2001S99 with ISBN 1-58563-200-7. The transcripts for this publication are available as Speech in Noisy Environments (SPINE1) Training Transcripts LDC2000T49 and Speech in Noisy Environments (SPINE1) Evaluation Transcripts LDC2000T54.

This work was sponsored in part by National Science Foundation Grant No. IIS-9982201.

*Data*

For an example transcript, please click here. There are a total of 253 files, one "game" each, for a rough total of 19 hours and 28 minutes (~4.4Gb) of audio data.

This corpus will be used as part of the training set for the Second Speech in Noisy Environments Evaluation (SPINE2). SPINE2 will provide a continuing forum for assessing the state of the art and practice in speech recognition technology for noisy military environments and for exchanging information on innovative speech recognition technology in the context of fully implemented systems that perform realistic tasks. The evaluation will provide researchers, potential sponsors, and customers with a quantitative means to appreciate the strengths and weaknesses of the technologies, and the results reported on will invite customer interest in the potential utility of the technologies. More information on this evaluation is available here.

*Updates*

There are no updates at this time.

*Introduction*

SummBank 1.0 was produced by Linguistic Data Consortium (LDC) catalog number LDC2003T16 and ISBN 1-58563-274-0.

SummBank 1.0 contains the data created for the Summer 2001 Johns Hopkins Workshop which focused on text summarization in a cross-lingual information retrieval framework. For more information about the Johns Hopkins summer workshop on Text Summarization please visit its website. The goal of the corpus is to gather together a corpus of original documents and summaries which can be used as gold standards by the documents summarization community.

The source of the data consists of 18,147 aligned bilingual (Cantonese and English) article pairs from the Information Services Department of the Hong-Kong Special Administrative Region of the People's Republic of China, which were published by the LDC in 2000 as Hong Kong News Parallel Text.

*Data*

This corpus contains 40 news clusters in English and Chinese, 360 multi-document, human-written non-extractive summaries, and nearly two million single document and multi-document extracts created by automatic and manual methods. The summarizer that was reimplemented and upgraded during the workshop is called MEAD; updated versions of the software are available from the MEAD website.

This distribution includes roughly two million text files, totalling approximately 13GB uncompressed. The text files are encoded either as utf-8 for English or GB or Big-5 for Chinese.

*Updates*

Additional information, updates, bug fixes may be available on the SummBank website.

*Introduction*

Switchboard Cellular Part 1 Audio was developed by the Linguistic Data Consortium (LDC) and consists of approximately 109 hours of English telephone conversations collected by LDC between 1999-2000. The Switchboard cellular collection focused primarily on GSM cellular phone technology. The project's goal was to target 190 subjects balanced by gender and under varied environmental conditions to participate in (10+) five to six minute conversations on GSM cellular phones. The speech data was collected for research, development, and evaluation of automatic systems for speech-to-text conversion, talker identification, language identification and speech signal detection purposes.

During the study period, LDC collected a total of 1,309 calls, or 2,618 sides (1,957 GSM), from 254 participants (129 male speakers, 125 female speakers) under varied environmental conditions.

*Data*

This release contains speech data files with documentation describing speaker information (sex, age, education, city and state where raised), call information (date, time, call duration, Personal Identification Numbers, topic) and audit information (channel quality, background noise). The data files are not compressed. The documentation also contains reports on clipped files.

Each speech file consists of a 1,024-byte ASCII-formatted Sphere header, followed by two-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 LDC's telephone collection platform. The header also indicates the caller_pin, callee_pin, topic_id, cellular service/handset information and speaker demographic information.

Other releases in this series include:

Switchboard Cellular Part 1 Transcribed Audio (LDC2001S15)

Switchboard Cellular Part 1 Transcription (LDC2001T14)

Switchboard Cellular Part 2 Audio (LDC2004S07)

*Updates*

55 missing sphere files were added the corpus on August 29, 2012. All copies ordered after that date will include those files.

*Introduction*

Switchboard Cellular Part 1 Transcribed Audio was developed by the Linguistic Data Consortium (LDC) and consists of approximately 24 hours of English telephone conversations collected by LDC between 1999-2000. This release contains the speech data files that correspond to Switchboard Cellular Part 1 Transcription (LDC2001T14).

The full set of conversations (approximately 109 hours) from the Switchboard Part 1 study is available in Switchboard Cellular Part 1 Audio (LDC2001S13). Switchboard Cellular Part 2 Audio (LDC2004S07) contains approximately 200 hours of English telephone conversations collected by LDC in the Switchboard Part 2 study.

The Switchboard Part 1 cellular collection focused primarily on GSM cellular phone technology. The project's goal was to target 190 subjects balanced by gender and under varied environmental conditions to participate in (10+) five to six minute conversations on GSM cellular phones. The speech data was collected for research, development, and evaluation of automatic systems for speech-to-text conversion, talker identification, language identification and speech signal detection purposes.

*Data*

Each speech file consists of a 1,024-byte ASCII-formatted Sphere header, followed by two-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 LDC's telephone collection platform. The header also indicates the caller_pin, callee_pin, topic_id, cellular service/handset information and speaker demographic information. The documentation also contains reports on clipped files.

*Updates*

There are no updates at this time.