Mixer 3 Speech
|Item Name:||Mixer 3 Speech|
|Author(s):||Shudong Huang, Kevin Walker, David Graff|
|LDC Catalog No.:||LDC2023S02|
|Release Date:||March 15, 2023|
|Data Source(s):||telephone conversations|
|Project(s):||MIXER, NIST LRE, NIST SRE|
|Application(s):||language identification, speaker identification|
|Language(s):||English, Mandarin Chinese, Min Nan Chinese, Amharic, Bengali, Persian, Hindi, Italian, Japanese, Georgian, Khmer, Korean, Lao, Panjabi, Western Panjabi, Russian, Spanish, Tamil, Tagalog, Thai, Tigrinya, Urdu, Uzbek, Vietnamese, Wu Chinese|
|Language ID(s):||eng, cmn, nan, amh, ben, fas, hin, ita, jpn, kat, khm, kor, lao, pan, pnb, rus, spa, tam, tgl, tha, tir, urd, uzb, vie, wuu|
|Online Documentation:||LDC2023S02 Documents|
|Licensing Instructions:||Subscription & Standard Members, and Non-Members|
|Citation:||Huang, Shudong, Kevin Walker, and David Graff. Mixer 3 Speech LDC2023S02. Web Download. Philadelphia: Linguistic Data Consortium, 2023.|
Mixer 3 Speech was developed by the Linguistic Data Consortium (LDC) and comprises 3,200 hours of audio recordings of conversational telephone speech involving 3,875 speakers and 26 distinct languages. This material was collected by LDC from 2005-2007 as part of the Mixer project, and recordings in this corpus were used in NIST Speaker Recognition Evaluation (SRE) and NIST Language Recognition Evaluation (LRE) corpora, including 2006 SRE and 2007 LRE.
Researchers interested in applying those benchmark test sets should consult the respective NIST Evaluation Plans for guidelines on allowable training data for those tests. Data from 2006 SRE and 2007 LRE are available in the LDC Catalog: 2006 NIST Speaker Recognition Evaluation Training Set (LDC2011S09), 2006 NIST Speaker Recognition Evaluation Test Set Part 1 (LDC2011S10), 2006 NIST Speaker Recognition Evaluation Test Set Part 2 (LDC2012S01), 2007 NIST Language Recognition Evaluation Supplemental Training Set (LDC2009S05) and 2007 NIST Language Recognition Evaluation Test Set (LDC2009S04).
The audio recordings were generated using LDC's computer telephony system capable of collecting speech from the telephone network. Recruited speakers were connected through a robot operator to carry on casual conversations lasting up to 10 minutes. Subjects fluent in languages other than English were asked to complete at least one non-English call.
The documentation for this release contains information about the number of calls per subject and the number of calls per language. It also includes certain speaker demographic information, such as date of birth, level of education, native language, other language capability, place of birth, place of residence and occupation.
The Mixer 3 collection contains 19,595 telephone recordings. The raw digital audio content for each call side was captured as a separate channel, then merged to be presented as a 2-channel file; the files are formatted as 8kHz, 8000 samples/second, u-law encoded NIST SPHERE files.
Please listen to this audio sample.
None at this time.
Additional Licensing Instructions
This members-only corpus is available to current members. Contact firstname.lastname@example.org for information about becoming a member.