BOLT CTS CALLFRIEND CALLHOME Mainland Mandarin Chinese Audio
| Item Name: | BOLT CTS CALLFRIEND CALLHOME Mainland Mandarin Chinese Audio |
| Author(s): | Jennifer Tracey, David Graff, Song Chen, Stephanie Strassel |
| LDC Catalog No.: | LDC2025S04 |
| ISLRN: | 319-237-059-416-8 |
| DOI: | https://doi.org/10.35111/d6y0-25s04 |
| Release Date: | May 15, 2025 |
| Member Year(s): | 2025 |
| DCMI Type(s): | Sound |
| Sample Type: | 16-bit |
| Sample Rate: | 8000 |
| Data Source(s): | telephone conversations, telephone speech |
| Project(s): | BOLT |
| Application(s): | cross-lingual information retrieval, information retrieval, machine translation, speaker identification |
| Language(s): | Mandarin Chinese |
| Language ID(s): | cmn |
| License(s): |
LDC User Agreement for Non-Members |
| Licensing Instructions: | Subscription & Standard Members, and Non-Members |
| Citation: | Tracey, Jennifer, et al. BOLT CTS CALLFRIEND CALLHOME Mainland Mandarin Chinese Audio LDC2025S04. Web Download. Philadelphia: Linguistic Data Consortium, 2025. |
| Related Works: | View |
Introduction
BOLT CTS CALLFRIEND CALLHOME Mainland Mandarin Chinese Audio was developed by the Linguistic Data Consortium (LDC) and consists of approximately 93 hours of speech from 236 unscripted telephone conversations between native speakers of the Mandarin Chinese dialect spoken in mainland China.
The DARPA BOLT (Broad Operational Language Translation) program developed machine translation and information retrieval for less formal genres, focusing particularly on user-generated content. LDC supported the BOLT program by collecting informal data sources -- discussion forums, conversational telephone speech, text messaging and chat -- in Chinese, Egyptian Arabic and English. The material in this release represents the unannotated Chinese source conversational telephone speech. The telephone data was transcribed, translated, and annotated for various tasks in the BOLT program including word alignment, treebanking, and co-reference.
CALLFRIEND and CALLHOME were collections of multilingual telephone speech conducted by LDC in support of language identification and speech identification technology development.
Data
The majority of the recordings in this corpus (141 calls, approximately 60%) are publicly released for the first time. The remainder (95 recordings, approximately 40%) were previously published in various CALLFRIEND, CALLHOME and HUB5 Mandarin datasets.
All calls originated in North America and were placed to locations overseas. Most participants called family members or close friends. Participants spoke on topics of their choice. Conversations lasted up to 30 minutes. Completed calls passed through a human auditing process to verify that the target language was spoken by the participants, to check the quality of the recordings, and to record information about dialect, noise and distortion.
The data is divided into training, development, and evaluation partitions as follows:
| Partition | Audio Files | Total Hours |
| train | 170 | 71.28 |
| dev | 18 | 4.12 |
| eval | 48 | 18.48 |
| total | 236 | 93.88 |
Audio files are presented in FLAC-compressed MS-WAV/RIFF format; they are two-channel, 16-bit sample data (converted from original mu-law samples) at 8000 samples/second. The audio is contained in two interleaved channels per file representing the two independently recorded sides of the telephone conversation.
Samples
Please listen to this speech sample (.flac)
Acknowledgement
This material is based upon work supported by the Defense Advanced Research Projects Agency (DARPA) under Contract No. HR0011-11-C-0145. The content does not necessarily reflect the position or the policy of the Government, and no official endorsement should be inferred.
Updates
There are no updates at this time.