Skip to content

Licence and citation

Licence

The clips are cut from recordings in TalkBank, whose data are shared under CC BY-NC-SA 3.0 unless a corpus says otherwise. As an adaptation, this dataset is shared under the same terms in their current version, CC BY-NC-SA 4.0:

  • non-commercial: no use in commercial products or services, including commercial models;
  • attribution: cite the dataset and the source corpora (below);
  • share-alike: anything you build from the clips and redistribute carries the same licence.

The code in scripts/ is Apache-2.0.

TalkBank ground rules

Using these clips means accepting TalkBank's ground rules and code of ethics as well. In short:

  • cite MacWhinney (2000) and the reference of every corpus you use;
  • respect the participants' privacy: do not try to identify the children or their families, and stop working with the data if you recognise someone who has not given you permission;
  • do not upload the clips to online services that keep the data, such as AI tools that train on what they are given, unless you have chosen an option that does not store it.

Citation

Cite the dataset:

@misc{karaca2026mapada,
  author       = {Karaca, H{\"u}seyin},
  title        = {{MaPaDa}: Children's \emph{mama}, \emph{papa} and \emph{dada} for Keyword Spotting},
  year         = {2026},
  version      = {1.0.0},
  publisher    = {Zenodo},
  doi          = {10.5281/zenodo.23642275},
  howpublished = {\url{https://hkaraca.com/mapada/}}
}

and, as TalkBank requires, the general CHILDES reference, the PhonBank reference if you use the French clips, and one reference for every corpus you use. All of them are in metadata/references.bib:

  • MacWhinney, B. (2000). The CHILDES Project: Tools for Analyzing Talk (3rd ed.). Lawrence Erlbaum Associates.
  • Rose, Y., & MacWhinney, B. (2014). The PhonBank Project. In The Oxford Handbook of Corpus Phonology (pp. 380–401). Oxford University Press.

Please also acknowledge the grants that support the databases: NICHD HD082736 (CHILDES) and NIH-NICHHD RO1-HD051698 (PhonBank).

Source corpora

collection corpus DOI cite clips
Eng-NA Bloom 10.21415/T50W4J bloom1974imitation 42
Eng-NA Brent 10.21415/T5MG74 brent2001role 310
Eng-NA Gleason 10.21415/T5101R bellinger1982sex 13
Eng-NA MacWhinney 10.21415/T5JP4F macwhinney1991childes 12
Eng-NA Sachs 10.21415/T5460C sachs1983talking 5
Eng-NA Snow 10.21415/T5C59M macwhinney1990childes 8
Eng-NA Soderstrom 10.21415/T5J595 soderstrom2008acoustical 1
Eng-NA VanHouten 10.21415/T5Z014 vanhouten1986role 2
Eng-NA Weist 10.21415/T5KW3F weist2009finiteness 29
Eng-UK Edinburgh 10.21415/SF64-JK03 ota2018choochoo 29
Eng-UK Forrester 10.21415/T5N31C forrester2002appropriating 2
Eng-UK MPI-EVA-Manchester 10.21415/T5DW48 lieven2009twoyearold 485
Eng-UK Sekali 10.21415/GB53-5W89 morgenstern2018blossoming 10
Eng-UK Thomas 10.21415/T5JG64 lieven2009twoyearold 6
French GoadRose 10.21415/T59P5N rose2000headedness 79
French KernFrench 10.21415/T5Q88C kern2009babbling 391
Eng-NA Braunwald 10.21415/T5D89Z braunwald1971mother timestamps only
Eng-NA Nelson 10.21415/T56W4D nelson1989narratives timestamps only

Two corpora ask for more than attribution. The Soderstrom corpus's owner asks to be sent a copy of any peer-reviewed publication that uses it. Braunwald and Nelson are below.

Held-back corpora

  • Braunwald. The corpus's author asks that any sound file used publicly, including in an online publication, be submitted to her first, so the 101 Braunwald utterances are included only as time stamps.
  • Nelson. The corpus asks that permission be obtained before publication. Its 6 utterances are included only as time stamps.

Their time stamps are in metadata/excluded_utterances.csv. If you have TalkBank access and the corpus owner's permission where it is needed, you can rebuild them with rebuild_from_childes.py.

Removal requests

If you contributed one of the source corpora, are a participant or a participant's family, or represent TalkBank and want clips removed, open an issue or email huseyin.karaca@bilkent.edu.tr. Clips will be removed in the next version, and the current version will be withdrawn where the host allows it.