Speed-Tb Phase 1 Nyishi Narration

About

About Nyishi

Nyishi is an under-resourced Tibeto-Burman language spoken by the Nyishi people, the largest ethnic group in Arunachal Pradesh, India. According to Census 2011, there are approximately 3 lakh speakers of the languages. The language belongs to the Tani branch of the Tibeto-Burman language family. Owing to its status as one of the largest languages of Arunachal Pradesh, some resources for the language development such as corpus and a language model have been developed. However, the language lacks any large significant speech or text corpus.

Dataset Description

The Nyishi Speech Dataset, developed as part of the Speech Datasets and Models for Tibeto-Burman Languages (Project SpeeD-TB) funded under Mission Bhashini, is a transcribed speech corpus of Nyishi,. The full dataset comprises over 200 hours of high-quality audio recordings paired with accurate transcriptions in both IPA and Roman script, making it the the largest speech resource for the language that not only enables building and evaluating voice models in low-resource linguistic settings but also enables large-scale linguistic description and documentation of the language. The audio data captures a diverse range of speakers across different age groups, genders, and education, ensuring variability in pronunciation, speech patterns, and tone. It includes both spontaneous and read speech collected in naturalistic and semi-controlled environments, thereby reflecting real-world linguistic usage. The transcriptions are carefully prepared and normalised to maintain consistency, supporting robust model training. This dataset also contributes to the preservation and digital documentation of the Nyishi language and culture by transforming oral knowledge into structured, machine-readable formats.

Almost 60% of the data in the corpus is included from domains of agriculture, education and science & technology. Rest of the data is from varied domains including culture, lifecycle, sports, entertainment, healthcare and oral history, thereby, giving a large coverage. We have also used a variety of elicitation methods for collecting the data including translations, narrations, lectures, role-play, spontaneous conversations, interviews and picture and video descriptions. The released dataset is meticulously mapped to a rich metadats including demographic and linguistic metadata of the speakers, domains, elicitation methods and to individual prompts. The audio included in the current dataset is already sliced at sentence level, thereby, ready to be integrated into the model training pipeline out-of-the-box.

The overall dataset of the project is collected over multiple phases and using multiple questionnaires. This repository contains the full dataset for the language collected till now.

Ethical Considerations, IPR and Attribution

This repository represents our committment to not only fair remuneration to the speakers of the language but also to co-ownership and equal IPR to all the contributors who have built the dataset. We believe this is the first step to move away from the extractive data collection and use practices and ensure fairness in our treatment of the community members. As such, we have listed all speakers and transcribers as Contributors to the dataset (we insist that they are co-owners of the dataset even though HuggingFace does not provide us an explicit way of stating that) and they are further recognised as Speakers and Annotators of the dataset. This dataset is only licensed to other researchers for use in their research projects. More details about licensing and commercial use conditions are given in License and Commercial Use sections.

Tools

We employed Karya and Atekho for collecting and recording data. The complete dataset is transcribed and exported using MATra Lab. Both Atekho and MAtra Lab are part of the LiFE Suite Ecosystem, developed by Unreal Tece LLP.

Speakers

Yana Banang, Toko Ipa, Likha Rich, Dorai Eje, Ichiko Pito, Tagan Mage, Likha Ya Joram, Joram Tuka, Joram Motu, Toko Doni, Joram Puj, Joram Maka, Likha Ana , Joram Maloti, Joram Nama and Nabam Runaya, Licha Ribia, Toko Yaram, Joram Nikam and Joram Chobin, Lishi Taza Joram, Joram Paul and Joram Yama, Joram Tane, Totosuka, Pachurinya, Joram Tapu, Tarak Tatam, Taj Takam, Pochu Mangu, Likha Tana, Joram I, Nikh Tado, Nangram Tare, Nabam Jakap, Mina Nihnaka, Taj Jarbo, Likha Ramleo, Taj Jikare, Joram Tahe, Dobiam Aniya, Joram Tami, Joram Niya, Joram Aina and Joram Dui, Pel Kamin, Joram Jyoti, Joram Byani, Hinium Mama, Pel Kamin, Joram Obbi, Joram Chobin, Likka Aku, Joram Yano, pachu mello, LIKHA BAI, Meko Singhi, Lishi Aku, Likha Aku, Taj Jarbo, joram paul, Joram Nikam , Baby Mugli , Tare Yaya, Tar Ado, honi japan, Talomary, DOBIAM TAKA, taho taku, Techi mary, Joram Tangam, Shri Joram Taja, Talo Paga , byabang chingsu , toko laam, TaloMartha, Pochu, Bagjam ana, Aiya gollom, Biri Yachu , Byabang sime , Toko rina, toko lel, Techi Kujma, Neelam tolu, phil tedi, donu kama, Tania Tayer , Byabang Yamin , DUI TAKA, jorambell , Joram Taja, Deadpool, Bengia Ania, lod yayo , Joram Aku, tame , yowa Yapin , talo badal, uku toko, Neelam Joshua , biki sima, phill doni, NABAM KHANDU

Annotators

Ishita Chowdhury, AnaghaS, SpeeD-TB Project, ManashiM, Lishi Aku, Anindita, Adrita Bhattacharya, AnishaDutta

Structure

The dataset is organized by splits (e.g. train, test, validation).

Each row contains audio, audio-level metadata, prompt metadata and speaker metadata as described below:

Audio and Audio-level Metadata

• audio: The audio file path (loaded as Audio feature in HF Datasets)
• audio_id: Unique identifier for the audio
• filename: Original filename
• sentence-<SCRIPT>-transcription: Text transcription of the audio in the given script
• speaker_id: Identifier for the speaker
• boundaryID: Identifier for the boundary
• start_time: Start time of the segment in seconds
• end_time: End time of the segment in seconds

Prompt Metadata

• 'Q_Id`: Unique identifier for the question or prompt associated with the audio (maps to the question in the LiFE Questionnaire projects and accessible through the questionnaire repo)
• 'Domain': Domain of the audio (e.g., Agriculture, Education, General, etc.)
• 'Elicitation_Method`: Method used to elicit the speech (e.g., Translation, Narration, etc.)
• Target: An optional field for translation indicating the grammatical structure being targeted for elicitation using the sentence.

Speaker Metadata

• ageGroup: Age group of the speaker (e.g., 18-30, 30-50, etc.)
• gender: Gender of the speaker
• educationLevel: Education level of the speaker
• educationMediumUpto12-list: Medium of education up to 12th grade (list of comma-separated values)
• 'educationMediumAfter12-list`: Medium of education after 12th grade (list of comma-separated values)
• otherLanguages-list: Languages spoken by the speaker (list of comma-separated values) - this usually excludes the primary language of the dataset and is used to capture multilingualism in speakers.
• nativeLanguage: The native language of the speaker (optional field if data is collected from non-native speakers of the language)
• placeOfRecording: The location where the audio was recorded (optional field) or the native place of the speaker (if known)
• typeOfplace: Whether the placeOfRecording mentioned is City, Town or Village.

Additional Metadata

• textgrid_json: TextGrid data converted to JSON format In addition to any other metadata fields provided during upload are optionally included.

License

This work is licensed under a CC-By-NC-SA-4.0 license. This license allows reusers to distribute, remix, adapt, build upon, and incorporate into software systems, the material in any medium or format for noncommercial purposes only, and only so long as attribution is given to the creator. If you remix, adapt, build upon, or incorporate into software systems, you must license the modified material, including material generated by the software system, under identical terms, and license the software system under the GNU General Public License.

Commercial Use

If you are interested in using this dataset for commercial purposes, please contact us (contact [at] unreal-tece[dot]co[dot]in). Profits from commercial licensing will be distributed as royalties to the community members who contributed to this dataset.

Contact

For questions, issues, or contributions, open an issue on the dataset repository or contact us directly (contact [at] unreal-tece[dot]co[dot]in).