drewoodward/miami-corpus

Bangor Miami Corpus (merged)

This repository contains a merged version of the Bangor Miami Corpus of Spanish–English bilingual speech:

• miamiCorpus_merged.mp3 — all 56 audio recordings concatenated into a single ~35-hour MP3 file.
• miamiCorpus_merged.vtt — the corresponding transcripts concatenated into a single WebVTT file.

This is not the canonical version. The canonical corpus (56 separate .wav / .cha file pairs in CHAT format, with gloss and translation tiers) is available from the original distributors:

• http://bangortalk.org.uk
• https://talkbank.org/biling/access/Bangor/Miami.html

If you need the per-conversation files, speaker metadata, CHAT tier annotations, or the authoritative version, please use the sources above rather than this mirror.

About the corpus

The Miami Corpus consists of 56 audio recordings and their corresponding transcripts of informal conversation between two or more speakers, involving 84 speakers living in Miami, Florida (USA). It was recorded and transcribed between 2008 and 2011 at the ESRC Centre for Research on Bilingualism at Bangor University, as part of a research project on code-switching. In total it consists of 242,475 words of text from 35 hours of recorded conversation.

License

This corpus is distributed under the GNU General Public License version 3 or later (https://www.gnu.org/licenses/gpl-3.0.html), following the terms set by the original distributors.

Important restriction on LLM training

The BangorTalk distributors state the following restriction:

Permission is NOT granted to use any of the material on this website to train an AI large language model UNLESS all the training data for that LLM is made publicly available.

If you intend to use this corpus for LLM training, you must comply with this condition. Downstream users of this Hugging Face mirror are equally bound by it.

TalkBank Code of Ethics

Researchers who use this corpus are requested to subscribe to the TalkBank Code of Ethics: https://talkbank.org/share/ethics.html

Citation

Please refer to the corpus as the Bangor Miami corpus and provide a link to one of the canonical websites (http://bangortalk.org.uk or http://www.talkbank.org).

Preferred reference:

Deuchar, M., Davies, P., Herring, J. R., Parafita Couto, M. C., & Carter, D. (2014). Building bilingual corpora: Welsh–English, Spanish–English and Spanish–Welsh. In E. M. Thomas & I. Mennen (Eds.), Advances in the Study of Bilingualism (pp. 93–110). Bristol: Multilingual Matters.

The corpus creators request that a copy of any publications that make use of this corpus be sent to:

Margaret Deuchar, ESRC Centre for Research on Bilingualism, Bangor University, Bangor, Gwynedd LL57 2DG, United Kingdom — m.deuchar@bangor.ac.uk

Privacy

Per the original documentation: names of speakers have been replaced by pseudonyms in the transcripts; utterances by non-consenting speakers have been transcribed as www and replaced by silence in the audio; and names associated with sensitive information have been beeped over. These protections are preserved in the merged version.

Acknowledgments

Corpus © 2010–present Margaret Deuchar, Maria del Carmen Parafita Couto, Peredur Webb-Davies, Kevin Donnelly, and the ESRC Centre for Research on Bilingualism at Bangor University. Recording and transcription funded by the UK Economic and Social Research Council.