Loading datasets...
Release Date: 9/17/2026
Format: MP3
Size: 220.08 MB
A collection of read speech recordings in Dameli (Dameli).
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.
dml)This datasheet is for cv-corpus-27.0-2026-09-11 of the Mozilla Common Voice Scripted Speech dataset for Dameli [Dameli - dml]. The dataset contains 6482 clips representing 10.83 hours of recorded speech (10.22 hours validated) from 5 speakers, recorded from a text corpus of 6,346 sentences.
Dameli is one of the most vulnerable languages of Pakistan. The language is spoken in a few remote villages, Asper, Dondidari, Ponagram and Shintari and the surrounding hamlets in the side valley called Damel in northern mountainous area of district Chitral of Khyber Pakhtunkhwa province. This vulnerability becomes more critical because of the community’s fewer numbers of speakers (about 6500 in total) In UNESCO’S Atlas of the world languages in Danger, Dameli is listed as “Severely endangered” (Elnazarov, 2010).The entry on Dameli was contributed by Hakim Elnazarov, and was based on information in Decker (1992).
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 | - | - |
| female_feminine | Female, feminine | - | - |
| transgender | Transgender | - | - |
| non-binary | Non-binary | - | - |
| do_not_wish_to_say | Prefer not to say | - | - |
| - | Unspecified | 6,482 (100.0%) | 5 (100.0%) |
Gender declared: 0 of 6,482 clips (0.0%), 0 of 5 speakers (0.0%)
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 | 614 (9.5%) | 1 (20.0%) |
| thirties | Thirties | - | - |
| fourties | Fourties | 5,129 (79.1%) | 2 (40.0%) |
| fifties | Fifties | 226 (3.5%) | 1 (20.0%) |
| sixties | Sixties | - | - |
| seventies | Seventies | - | - |
| eighties | Eighties | - | - |
| nineties | Nineties | - | - |
| - | Unspecified | 513 (7.9%) | 5 (100.0%) |
Age declared: 5,969 of 6,482 clips (92.1%), 0 of 5 speakers (0.0%)
Clip buckets
| Bucket | Clips |
|---|---|
| Validated | 6,115 (94.3%) |
| Invalidated | 68 (1.0%) |
| Other | 299 (4.6%) |
Training splits
| Split | Clips |
|---|---|
| Train | 3,190 (52.2%) |
| Dev | 1,101 (18.0%) |
| Test | 626 (10.2%) |
Training split coverage: 4,917 of 6,115 validated clips (80.4%)
The dataset contains 6115 validated, 68 invalidated, and 299 unresolved clips. The average clip duration is 6.019 seconds.
The corpus consists of 5,670 sentences in the Dameli language. The data was collected from multiple sources, including published books in Dameli, community-written materials, and newly created sentences designed to reflect everyday use of the language. The aim of compiling this corpus is to represent a wide range of topics such as social life, education, agriculture, economy, poetry, farming, and history. This balanced collection provides a valuable resource for linguistic analysis, documentation, and language technology development.
Validated sentences: 5,670
| Category | Count |
|---|---|
| Unvalidated sentences | 676 |
| Pending sentences | 676 |
| Rejected sentences | - |
| Reported sentences | - |
The corpus contains 6,346 sentences: 5,670 validated and 676 unvalidated (676 pending review, 0 rejected), with 0 reported for review.
The Dameli corpus is written using the Arabic script (Perso-Arabic style), which is commonly used for many regional languages in Pakistan. The writing system has been adapted to represent Dameli sounds, with some additional diacritics and letters used where necessary to capture specific phonetic distinctions.
آ ا ب پ ت ٹ ث ج چ ڇ څ ح خ د ڈ ذ ر ڑ ز ڙ ژ س ش ݜ ص ض ط ظ ع غ ف ق ک گ ل م ن ݨ و ہ ھ ء ی ے
There follows a randomly selected sample of five sentences from the corpus.
ماں لاکی اپھامے ییسن بہ تسہ کی کُرئی کُھنڈِن
کلونہ ڇھاریسن غلامی ݜا غلامیہ
تو ساں یین آڅنہ ائی ایس ماڜ ساں نظرہ آگیم
تس تہ پوچئے تے تاں مرگ کیہ شئیہ درو
گرم تہ اِک کتیِکی ژیڜٹیرانم تساں کُتہ گین تنی گنے تے ترہ ژا ماں کیہ نی اُوتھی درو تے ائی گیرنکھتی آڅیم
The text corpus was compiled from the following sources:
Published books in the Dameli language
Unpublished community manuscripts and notes
Folk stories, oral traditions, and poetry transcribed into written form
Newly created sentences for grammar and vocabulary coverage
Educational and social materials produced by local speakers
| Source | Sentences |
|---|---|
| زادی تہ کاݨ | 2,429 (42.8%) |
| خاکسار شاعری | 1,722 (30.4%) |
| دامیاں باڜہ تہ شُلوک | 886 (15.6%) |
| Dameli Proverbs | 286 (5.0%) |
| دمیلی،اُردو،انگریزی بول چال | 208 (3.7%) |
| عصمت اللہ دمیلی (زادی تہ کاݨ) | 100 (1.8%) |
| Other | 38 (0.7%) |
General, Agriculture and Food, Finance, Service and Retail, Healthcare, History, Law and Governmant, Media and Entertainment, Nature and Environment, Language Fundamentals (e.g. Digits, Letters, Money)
| Code | Domain | Clips | Speakers |
|---|---|---|---|
| general | General | 7 (0.1%) | 3 (60.0%) |
| agriculture_food | Agriculture and Food | - | - |
| automotive_transport | Automotive and Transport | - | - |
| finance | Finance | - | - |
| service_retail | Service and Retail | - | - |
| healthcare | Healthcare | - | - |
| history_law_government | History, Law and Government | 3 (0.0%) | 2 (40.0%) |
| media_entertainment | Media and Entertainment | - | - |
| nature_environment | Nature and Environment | 4 (0.1%) | 2 (40.0%) |
| news_current_affairs | News and Current Affairs | - | - |
| technology_robotics | Technology and Robotics | - | - |
| language_fundamentals | Language Fundamentals | - | - |
The collected texts were first gathered from books, manuscripts, and community contributions. Additional sentences were created to cover gaps in vocabulary and grammar. All materials were transcribed into a consistent format using the Arabic (Perso-Arabic) script adapted for Dameli. The data was then carefully reviewed and proofread to remove errors and ensure accuracy. Finally, the sentences were digitized, standardized, and compiled into a single corpus of 5,670 sentences for use in research and language development.
Users of this dataset may consider the following post-processing steps depending on their research goals:
Normalization: Ensure consistent spelling, especially where multiple variants of the same word exist.
Tokenization: Segment the text into words or morphemes for computational use.
POS tagging / annotation: Add part-of-speech or grammatical tags if the dataset will be used for linguistic or NLP applications.
Transliteration: Convert the Arabic script into Latin script if required for cross-linguistic comparison.
Alignment: If paired with translations, align Dameli sentences with their equivalents in other languages for bilingual analysis.
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
sentence - the sentence to be read aloud
sentence_id - unique identifier for the sentence
sentence_domain - domain classification(s) of the sentence
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
locale - locale code of the language
segment - if sentence belongs to a custom dataset segment, it will be listed here
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)
As internet access is limited in the Dameli Valley, most local communication takes place through community gatherings, cultural events, and village meeting. However, Dameli people living in cities and outside the valley stay connected online. They maintain a WhatsApp group called “Anjuman Taraqi Damyan Basha”, where members share poetry, cultural materials, news, and language-related resources. In this way, both offline and online platforms help keep the community connected and engaged in language preservation.
There are no formal online forums or blogs for discussions related to the dataset. Instead, most of the discussion and coordination took place in the WhatsApp group “Anjuman Taraqi Damyan Basha”, where community members exchanged ideas, shared poetry, cultural materials, and contributed to decisions during the dataset creation process.
Common Voice Community
This dataset was partially funded by the Open Multilingual Speech Fund managed by Mozilla Common Voice.
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