Common Voice Scripted Speech 25.0 - Tush

ვაჲღეჼ — Tush (bbl)

This datasheet is for cv-corpus-25.0-2026-03-09 of the Mozilla Common Voice Scripted Speech dataset for Tush [ვაჲღეჼ - bbl]. The dataset contains 4982 clips representing 12.16 hours of recorded speech (11.17 hours validated) from 23 speakers, recorded from a text corpus of 2,241 sentences.

Language

This language belongs to the Nakhi language family and is currently endangered. The language is acknowledged by researchers as much older than the two other languages of this language family - Chechen and Ingush. The grammar of this language was the first to be scientifically studied among all Iberian-Caucasian languages (A. Schiefner, Versuch über die Tush-Sprache, Petersburg, 1854). This ancient language, brought to this day by the ethnically Georgian (Orthodox Christian) aboriginal bilingual population, was spoken by more than 3,000 people a century ago. This language was known, and still is known, in the main settlement of the Tush people - the village of Zemo Alvani - by people from other parts of Georgia and other nationalities who settled there during World War I-II. In recent decades, as a result of the outflow of the working-age population from the country, the connection between generations has been severed, and mixed marriages have become more common. That's why today only the older generation knows the language well, the 40+ generation not so well, and those under 25 mostly only understand it and cannot speak. That is, fewer than 800 people scattered around the world speak this language to a greater or lesser extent, and about 400 locally. In fact, the scientific study of the language today is in complete stagnation - the language is unexplored. Although there is a lot to be studied, both in terms of this language itself and its connection with the Georgian-Kartvelian and Hurrian languages. This language course is no longer taught in the country's higher education institutions. The language is referred to by different terms - the Georgian term Tushuri and the Russian term тушинский; during the USSR, it was replaced by the term Batsburi - бацбийский, and after that by Tsova Tush. People are dissatisfied - by changing the name of their language, the Tushetians - this border-defending people, praised many times in history - appear as an unclear ethnicity. But the language is still alive - enthusiasts even translate poetry into this language, while maintaining rhyme, meter, and rhythm.

Demographic information

The dataset includes the following self-declared age and gender distributions. A coverage summary is shown below each table.

Gender

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.

Gender declared: 3,243 of 4,982 clips (65.1%), 7 of 23 speakers (30.4%)

Age

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.

Age declared: 4,706 of 4,982 clips (94.5%), 19 of 23 speakers (82.6%)

Data splits for modelling

Clip buckets

Training splits

Training split coverage: 1,161 of 4,580 validated clips (25.3%)

The dataset contains 4580 validated, 130 invalidated, and 272 unresolved clips. The average clip duration is 8.787 seconds.

Text corpus

Presently, the corpus contains single sentences or texts with a few sentences (5 to 30 sentences). The average length of the sentences is from 8 to 15 words. These texts are created by enthusiasts specifically for Common Voice.

Validated sentences: 1,661

The corpus contains 2,241 sentences: 1,661 validated and 580 unvalidated (573 pending review, 7 rejected), with 0 reported for review.

Writing system

Georgian alphabet with additional signs, symbols.

Symbol table

ა ბ გ დ ე ვ ზ თ ი ჲ კ ლ ლ' მ ნ ჼ ო პ ჟ რ ს ტ უ უ̂ ფ ქ ღ ყ შ ჩ ც ძ წ ჭ ხ ჴ ჯ ჰ ჰ̦ ჵ ჸ ჺ ა́ ე́ ი́ ო́ უ́ ე̆ ი̆ ო̆ უ̆

Sample

There follows a randomly selected sample of five sentences from the corpus.

1. თურ ქირჯოლა́ს, მო́ჸ ლივა́ს, იშტიკ, ქიბალა სო́ჼ მე́ წყეგეჸ ჲარსტენო̆ ჲაე́ ნა́ხლო ჩე́ვარლა.

2. მეზო́ბლი დარ ბა́რლეჼ ლო́უ̆მო̆ დუ́ჲჰ̦რელუშ კა́რვივ ჰ̦ალო̆ და́ხკენო̆ ყო́ნში, მახკარ-კნა́თი.

3. კაწ აჩე́ნადალინი̆ სე ე́ზო სიფსრე, შინ ბუჲსო̆ სეჲჩუ და́ხერ, ნი́დაგ თე́ლ'რა́ს, გო́ჩნადორა́ს.

4. ინც დითხოღეჼ ვო́მაჸ ჩუ დე́ჸენო̆ და, ჰ̦ა́ნნ ხეჸ მაცა́ხლუჼ, მო́ლუჩო დაჴდალინოე́ უ̂ხუშ თე́ლ'ინო̆.

5. შილღეჼ ვოჰ̦ ვარ სკო́ლე, ბჵა́ ჺა́რჭიჼ ბაქ ბაჲნო̆, ჴშირუშ კორ ბა́ყორ შხრამ-შხრამაჲნო̆.

Sources

Text domains

General Agriculture and Food Nature and Environment

Processing

Because there is no single agreed-upon font, we could not use texts copied from books. To compose sentences for the corpus, we developed a font that is as convenient as possible—simplified on the one hand, and refined with the addition of stressed vowels on the other. The material collected to date consists of short episodes written specifically for this corpus by several people. Editing was expressed in shortening sentences and reducing them to fewer than 15 words. Sometimes we had to check and clarify texts over the phone or in person. We also recorded the stories of those who knew the language best. The biggest obstacle turned out to be: 1. In Pirago, there is no single consonant () grapheme, for which we used the letter (ჰ̦), and the symbols for short vowels and consonants, placing them one by one on the graphemes, take a lot of time when typing; 2. With two exceptions, elderly people cannot independently create voice recordings due to their lack of computer skills, even though they know the language well.

Fields

Clips

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.tsv

The 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.tsv

The 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)

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

• Speak

• Write

• Listen

• Review

Acknowledgements

Datasheet authors

• Tinatin Tsiskarishvili

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 ↩3 ↩4