Common Voice Spontaneous Speech 3.0 - Arvanitika

aat — Arvanitika (aat)

This datasheet is for sps-corpus-3.0-2026-03-09 of the Mozilla Common Voice Spontaneous Speech dataset for Arvanitika [aat - aat]. The dataset contains 334 clips representing 2.3 hours of recorded speech (0 hours validated) from 5 speakers.

Language

Arvanitika (/ˌɑːrvəˈnɪtɪkə/; Arvanitika: αρbε̰ρίσ̈τ, romanized: arbërisht; Greek: αρβανίτικα, romanized: arvanítika), also known as Arvanitic, is the variety of Albanian traditionally spoken by the Arvanites, a population group in Greece. Arvanitika was brought to Southern Greece during the late Middle Ages by Albanian settlers who moved south from their homeland in present-day Albania in several waves. The dialect preserves elements of medieval Albanian, while also being significantly influenced by the Greek language. Arvanitika is today endangered, as its speakers have been shifting to the use of Greek and most younger members of the community no longer speak it.

Data splits for modelling

The dataset clips are categorised by transcription status and training-set assignment. The following tables summarise the distribution.

Audio clips

Training splits

Training split coverage: 0 of 0 transcribed & validated clips (0.0%)

Transcriptions

Transcription status

Writing system

Following the writing rules of modern Albanian, while keeping the Greek alphabet for sounds not existing in Albanian.

In cases of code-switching, write the Greek portion using Greek letters, even for isolated Greek words.

• Use the capital letter E to represent the Albanian letter ë [ǝ].

• Use the capital letter C to represent the Albanian letter ç [tš].

• For γ before α and ο, retain γ or alternatively write gh.

• For χ before ε and ι, retain χ or alternatively write hj

Samples

Questions

There follows a randomly selected sample of questions used in the corpus.

1. Çë bënje çë nuku keshe bën ikojenië tënde akoma?

2. Me shënden je mir?

3. Çë smund ke shkuar më shum?

4. Çë pun të pëlqen më shum të bësh?

5. Çë urat doe të jepnje djeltë që do rriteshin?

Responses

There follows a randomly selected sample of transcribed responses from the corpus.

1. ...anw e Papandis, ngru vajzo tw vesh nw klish, ra kambana di tri her, ngru vajzo tw vwsh Cemberw, ra kambana me ata die, ngru vajzo tw vwsh mandil nw krie

2. dhjavasmw efimeridh, dhjavasmw filladh, vemi peripato ... me shoqet, qeshm, thomi astia...

3. Faje qw mw pwlqen mw shum isht qw bwnet stifadhoa, me qep, me hudhrw, me tuti mirodhitw bwrdha, shter edhe dhafne bwrdha dhe dhafne bwrdha, nga tuti tewmirat, bwnej shum e mirw stifadho atje, se keshwm edhe mish tw mirw, mishi ish nga dhwntw twnw, nga dhitw tanw mishi, nuku, ish i mirw, edhe stifadho me kto, me ksidh, me uthul bwrdha, me tuti tw mirodhit, me tuti tw mira.

4. Cw jesh e vogwl, vej ndw jitoni, e doj t'i thosh Cw bwnjwnw, tw mw thojnw ngato punwtw ... gjigjesh Cw thojnw ato... mw pwlqejnw mua atw qw thojnw ato? Nuku mw pwlqejn, do iki, mw pwlqejn, do ri.

5. do flerw, e tha tw vwmw, tw shtitim kush do vejw njora atje, an do marrw atje qw ka rush tw ham, e thanw ne, vate edhe skanxohjeri, u nisw dhelpra tw vej mwnjora, skanxohjeri ... fjeti nw malw dhelprwsw bishtin, s'e muar mirw..., vate atje, vate atje qw kejnw bwnw simadhin se mos kush do vejw mwnjora edhe....

Fields

Each row of a tsv file represents a single audio clip, and contains the following information:

• client_id - hashed UUID of a given user

• audio_id - numeric id for audio file

• audio_file - audio file name

• duration_ms - duration of audio in milliseconds

• prompt_id - numeric id for prompt

• prompt - question for user

• transcription - transcription of the audio response

• votes - number of people that who approved a given transcript

• age - age of the speaker1

• gender - gender of the speaker1

• language - language name

• split - for data modelling, which subset of the data does this clip pertain to

• char_per_sec - how many characters of transcription per second of audio

• quality_tags - some automated assessment of the transcription--audio pair, separated by |

  • transcription-length - character per second under 3 characters per second

  • speech-rate - characters per second over 30 characters per second

  • short-audio - audio length under 2 seconds

  • long-audio - audio length over 5 minutes

Get involved

Community links

• Common Voice translators on Pontoon

• Common Voice Communities

Discussions

• Common Voice on Matrix

• Common Voice on Discourse

• Common Voice on Discord

• Common Voice on Telegram

Contribute

• Contribute questions

• Validate questions

• Answer questions

• Transcribe recordings

• Validate transcriptions

Acknowledgements

Datasheet authors

• Vivian Stamou <vivianstamou@gmail.com>

Funding

This dataset was partially funded by the Open Multilingual Speech Fund managed by Mozilla Common Voice.

Licence

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.

Footnotes

1. 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