REMIX Telephone Collection was developed by the Linguistic Data Consortium (LDC) and contains 320 hours of English conversational telephone speech from 358 speakers who had completed all tasks in one of the previous LDC Mixer collections, specifically, Mixers 4-7. The data was collected in 2012; recordings in this corpus were used to support the NIST 2012 Speaker Recognition Evaluation. 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 on suggested topics lasting up to 10 minutes. Subjects were asked to complete 12 calls, half of those in a "noisy" environment. Examples of proposed noisy environments included using a speakerphone, calling from a busy street, noisy store or office, or calling from a room with loud background noise. The documentation for this release includes call topics, the number of calls per subject, the number of noisy calls and certain speaker demographic information (e.g., year of birth, education level, occupation). The REMIX collection contains 1917 telephone recordings. The files are formatted as 2-channel, 8-bit, mu-law encoded sample data recorded at 8000 samples/second, with a NIST SPHERE-format header on each file.

Creator(s)

David Graff

Karen Jones

Stephanie Strassel

Kevin Walker

Distributor(s)

Linguistic Data Consortium

Right Holder(s)

Portions © 2012, 2023 Trustees of the University of Pennsylvania