language:
- ikk
license:
- apache-2.0
tags:
- audio
- speech
- religious texts
- low-resource languages
- niger-congo
- readings
ikk_bible_JHNandMRK_chapter1_to_10 Dataset
Dataset Description
Disclaimer
This dataset is not fully setup, you will encounter errors while trying to load the data.
Overview
The Ikaaudio dataset comprises segments from the first 10 chapters of the Ika translation of the New Testament. It contains verse-level audio segments, manually verified to include only high-quality segments were the transcription sufficiently matches the audio. The MMS_FA MMS Force Aligner, open-sourced by Meta AI research lab was used to generate segment the chapters of John and Mark. Upon segmentation, the alignments of the first 10 chapters of John and Mark were manually examined through listening and comparing audio to the text. Only quality segments and alignments where selected. The creation of this dataset was made possible by utilizing the publicly available transcripts and audio of the Ika new testament, courtsey of Faith Comes By Hearing, through the efforts of ikabible.org.
Languages
Ika language (ISO 639-3 code: ikk), translated from English and recorded in Ika.
Annotations
Annotation Process
The dataset was annotated through a process of translation from English to Ika, followed by reading and recording the Ika texts. This process ensured that the dataset not only captured the linguistic nuances of the Ika language but also maintained the integrity of the original biblical texts.
Who are the annotators?
The annotation and recording were conducted by personnel affiliated with ikabible.org, sponsored by Faith Comes By Hearing.
Personal and Sensitive Information
The dataset does not contain any personally identifiable information. However, users should note that the dataset comprises religious texts, specifically from the New Testament, which might be considered sensitive by some. The primary speaker in the recordings is male, which users should consider when utilizing the dataset to avoid potential gender bias in applications.
Considerations for Using the Data
Social Impact of Dataset
This dataset contributes to the digital representation of low-resource languages, specifically Ika, by providing high-quality audio recordings for language model training. While it serves a valuable linguistic and cultural preservation purpose, users should be mindful of the religious nature of the content and the potential sensitivity surrounding it.
Discussion of Biases
The dataset predominantly features a male speaker and contains religious texts, which could introduce specific biases. Users are encouraged to consider these factors in their applications to ensure inclusive and respectful use.
Other Known Limitations
The dataset has not been divided into training and test sets, which users may need to do manually to evaluate model performance effectively.
Additional Information
Dataset Curators
The dataset was curated with great care to verify the alignments of all segments and transcripts. If you utilize this dataset in your research or application, please cite and reference the work accordingly.
Citation
Details for citation will be provided by the dataset creator.
Licensing Information
This dataset is distributed under the Apache License 2.0.