Dagbani-TTS-Dataset

Language

Dagbani (ISO 639-3: dag), also known as Dagbane or Dagomba, is a Gur language of the Niger-Congo family (specifically the Oti-Volta branch) spoken primarily in the Northern Region of Ghana, in the traditional Dagbon kingdom centred on the town of Yendi. It is the most widely spoken language of northern Ghana, with an estimated 3 to 4 million speakers, and functions as a regional lingua franca across the north of the country. Dagbani is typologically characterised by agglutinative morphology, subject-verb-object (SVO) word order, and a system of noun classes. The language has a rich tonal system in which tone is lexically and grammatically distinctive. Dagbani also exhibits nasal vowels, prenasalised consonants, and labialised and palatalised consonant clusters.

Variants

The recordings in this dataset were produced by a single speaker of Dagbani from Ghana. The variety recorded is representative of the standard written and broadcast norm of Dagbani as used in the Dagbon area. No sub-variety distinctions are encoded in the dataset.

Alphabet

The orthography used in the transcription of the audio recordings follows the standard Dagbani writing system as developed and standardised by the Dagbani Orthography Committee, in use for Dagbani publications, literacy materials, and broadcast media in Ghana. The orthography is based on the Latin alphabet and employs a set of conventions to represent sounds specific to Dagbani. Tonal distinctions, while present in the spoken language, are not systematically marked in the standard orthography. Dagbani makes use of several digraphs (e.g., gb, kp, ny, ng) to represent labial-velar stops and palatal/velar nasals. Additional special characters include ɛ (open mid-front vowel), ɔ (open mid-back vowel), and ŋ (velar nasal), which are standard in the orthography. Vowel nasalisation is indicated by a following n in certain environments. The result is an orthographic system that is widely used in Dagbani literacy and publishing, making the transcriptions in this dataset consistent with established written norms.

Source

This dataset was compiled as part of a Text-to-Speech data collection initiative for Dagbani. The textual source material consists of a range of texts in Dagbani, including educational, narrative, and informational content. The speaker read prepared passages drawn from these texts, providing naturalistic and culturally grounded speech data. Recordings were made across 16 sessions and subsequently curated, deduplicated, and aligned with their corresponding transcriptions.

Domain

The dataset consists of prompted read speech in Dagbani. The textual source material derives from a range of registers, including educational, narrative, and informational texts — representative of standard written Dagbani. The recordings offer good prosodic and lexical diversity — including varied sentence lengths, clause structures, and registers — making them suitable for TTS model training and ASR evaluation.

Size

Total size of MP3 audio: approx. 185 MB Total size of TSV mapping files: approx. 300 KB

Structure

This dataset comprises audio clips and audio/text mapping files organised across 16 recording sessions. There are 2,488 audio clips in MP3 format, totalling 4 hours, 2 minutes and 9 seconds of speech. The dataset includes 16 audio/text mapping files (mapping.tsv), each containing aligned audio/sentence pairs for the corresponding session, with 2,488 aligned pairs in total. Each TSV file contains the following fields: audio_filename, key, sentence, attempts.

Sample