Loading datasets...
Task: ASR
Release Date: 3/22/2026
Format: MP3
Size: 12.15 MB
Share
A collection of read speech recordings in Western Sierra Puebla Nahuatl (nhi).
Restrictions/Special Constraints
None provided.
Forbidden Usage
It is forbidden to attempt to determine the identity of speakers in the Common Voice datasets. It is forbidden to re-host or re-share this dataset.
Intended Use
This dataset is intended to be used for training and evaluating automatic speech recognition (ASR) models. It may also be used for applications relating to computer-aided language learning (CALL) and language or heritage revitalisation.
nhi)This datasheet is for cv-corpus-25.0-2026-03-09 of the Mozilla Common Voice Scripted Speech dataset for Western Sierra Puebla Nahuatl [nhi - nhi]. The dataset contains 427 clips representing 0.6 hours of recorded speech (0.05 hours validated) from 6 speakers, recorded from a text corpus of 769 sentences.
Western Sierra Puebla Nahuatl, alternatively Zacatlán-Ahuacatlán-Tepetzintla Nahuatl, is a variety of Nahuatl spoken in the Northwestern region of Puebla's Sierra Norte. A 2009 report from Mexico's National Institute of Indigenous Languages estimates approximately 17,000 speakers.
The language code is nhi. There is quite a bit of variation within the nhi variant, varying between
municipalities and communities. For example, some towns, near San Miguel Tenango, use "inverted prefixes" compared
to the rest of the Nahuatl-speaking area, (e.g. in-nihnimi instead of ni-nihnimi, "I walk").
The sentences in this corpus come from the nhi Universal Dependencies treebank, which is made up of samples
from all three municipalities, and a set of dictionary-style example sentences written by a speaker from the
Tepetzintla municipality.
The dataset includes the following self-declared age and gender distributions. A coverage summary is shown below each table.
Self-declared gender information. The table shows clip and speaker counts with percentages. Speakers who did not declare a gender are listed as Unspecified. A dash (-) indicates zero.
| Code | Gender | Clips | Speakers |
|---|---|---|---|
| male_masculine | Male, masculine | 35 (8.2%) | 1 (16.7%) |
| female_feminine | Female, feminine | 312 (73.1%) | 1 (16.7%) |
| transgender | Transgender | - | - |
| non-binary | Non-binary | - | - |
| do_not_wish_to_say | Prefer not to say | - | - |
| - | Unspecified | 80 (18.7%) | 5 (83.3%) |
Gender declared: 347 of 427 clips (81.3%), 1 of 6 speakers (16.7%)
Self-declared age information. The table shows clip and speaker counts with percentages. Speakers who did not declare an age are listed as Unspecified. A dash (-) indicates zero.
| Code | Age | Clips | Speakers |
|---|---|---|---|
| teens | Teens | - | - |
| twenties | Twenties | 312 (73.1%) | 1 (16.7%) |
| thirties | Thirties | 35 (8.2%) | 1 (16.7%) |
| fourties | Fourties | - | - |
| fifties | Fifties | - | - |
| sixties | Sixties | - | - |
| seventies | Seventies | - | - |
| eighties | Eighties | - | - |
| nineties | Nineties | - | - |
| - | Unspecified | 80 (18.7%) | 5 (83.3%) |
Age declared: 347 of 427 clips (81.3%), 1 of 6 speakers (16.7%)
Clip buckets
| Bucket | Clips |
|---|---|
| Validated | 40 (9.4%) |
| Invalidated | - |
| Other | 387 (90.6%) |
Training splits
| Split | Clips |
|---|---|
| Train | 24 (60.0%) |
| Dev | 10 (25.0%) |
| Test | 6 (15.0%) |
Training split coverage: 40 of 40 validated clips (100.0%)
The dataset contains 40 validated, 0 invalidated, and 387 unresolved clips. The average clip duration is 5.081 seconds.
The average length of validated sentences is 5.5 words (34 characters). The corpus contains numerous Spanish loanwords, calques, and code-switching. For example, the following sentence from the corpus contains the Spanish conjunction pero "but", Spanish preposition de "of", the subordinator hasta "until", a morphologically-adapted loanword oniquestudiaro (from estudiar "to study"), and a morphologically-adapted number, ocho "eight" (tiochoque "we are eight"). Due to long-standing language contact and bilingualism, Nahuatl commonly incorporates elements of Spanish, and this is well-represented in the text corpus.
Pero de nochten tlen tiochoque sa ye neh oniquestudiaro hasta cuando onipiyaya dieciocho años.
"But of all of us eight I was the only one who studied until when I was eighteen years old."
Validated sentences: 758
| Category | Count |
|---|---|
| Unvalidated sentences | 11 |
| Pending sentences | 11 |
| Rejected sentences | - |
| Reported sentences | - |
The corpus contains 769 sentences: 758 validated and 11 unvalidated (11 pending review, 0 rejected), with 0 reported for review.
Western Sierra Puebla Nahuatl is written in the Latin script. Specifically, the Common Voice corpus uses an orthographic norm defined by the Summer Institute of Linguistics in collaboration with the nhi-speaking community in San Miguel Tenango, Zacatlán, Puebla. See this Description of the nhi alphabet for more information.
In most cases, Spanish loans, even those with Nahuatl morphology, are written in standard Spanish orthography (the Nahuatl morphemes follow the Nahuatl orthography described here, though it is worth noting the similarities between the writing systems so that there are only rare cases where the orthography change is noticeable).
a b c ch cu d e f g h i j k l m n ñ o p qu r s t tl tz u v w x y
Note that of the alphabet listed above, b d f g j k ñ r v w are used for loanwords and some proper names only.
There follows a randomly selected sample of five sentences from the corpus.
Yotiquitiya para inpiluan o para incal.
In Carranza oquimpiyaya nisoldados.
In Ignacio yomotlaleh.
Neh nochipa onechpactaya non.
Neh oniyol ich se altipetl.
Subset of UD corpus (Public domain)
Individual sentences submitted by users through the Mozilla Common Voice interface (Public domain)
| Source | Sentences |
|---|---|
| Created by speaker for Common Voice | 401 (52.9%) |
| Sasaki | 308 (40.6%) |
| personal narrative | 36 (4.7%) |
| Other | 13 (1.7%) |
General
The original sentences from the UD treebank are written in varying orthographic norms. Prior to their inclusion into the Common Voice corpus, the orthgoraphy was converted to the SIL San Miguel Tenango orthography ("ilv") using a rule-based automatic converter from the py-elotl Python package, followed by some manual verification.
Each row of a tsv file represents a single audio clip, and contains the following information:
client_id - hashed UUID of a given user
path - relative path of the audio file
text - supposed transcription of the audio
up_votes - number of people who said audio matches the text
down_votes - number of people who said audio does not match text
age - age of the speaker1
gender - gender of the speaker1
accents - accents of the speaker1
variant - variant of the language1
segment - if sentence belongs to a custom dataset segment, it will be listed here
prompt_upvotes - number of upvotes the sentence prompt received
prompt_reports - number of reports the sentence prompt received
is_edited - whether the clip's transcription has been edited
validated_sentences.tsvThe validated_sentences.tsv file contains one row per validated sentence in the text corpus:
sentence_id - unique identifier for the sentence
sentence - the sentence text
variant - the variant of the language
sentence_domain - the domain(s) the sentence belongs to
source - the source the sentence was collected from
is_used - whether the sentence is still in circulation for recording
clips_count - number of clips recorded for this sentence
unvalidated_sentences.tsvThe unvalidated_sentences.tsv file contains one row per unvalidated sentence in the text corpus:
sentence_id - unique identifier for the sentence
sentence - the sentence text
variant - the variant of the language
sentence_domain - the domain(s) the sentence belongs to
source - the source the sentence was collected from
up_votes - number of upvotes the sentence received
down_votes - number of downvotes the sentence received
status - current status of the sentence (pending or rejected)
Robert A. Pugh
This dataset is released under the Creative Commons Zero (CC-0) licence. By downloading this data you agree to not determine the identity of speakers in the dataset.
For a full list of age, gender, and accent options, see the demographics spec. These will only be reported if the speaker opted in to provide that information. ↩ ↩2 ↩3 ↩4