A Kiswahili dataset
The dataset contains text and audio files.The dataset contains 1,570 text files,and 1,570 audio files.The text files were collected from Kiswahili newspapers, articles, short story excerpts, and novels. The dataset contains 23,487 Kiswahili words. The words were then transcribed into mean words of 14.96 per text file and saved in CSV format. Each text file was divided into three parts: unique id, transcribed words, and normalized words. A unique id is a number assigned to each text file. The transcribed words are the text spoken by a reader. Normalized texts are the expansion of abbreviations, numbers, and the letter-sequence into full words. A recording per line was made for each text file and assigned a unique id, same as the text file. The audio recording was made in the audacity software using a professional microphone in a least-noisy environment. The audio length for each recording lasted between 1s to 12.5s. .Each audio file's properties include a sampling rate of 22.05 kHz and 16-bit single-channel unsigned PCM. The audio and the text file created meet the required properties for use in building a TTS system.
Steps to reproduce
i. The text corpus was collected from novels, newspaper articles, and short story excerpts. ii. The text was then transcribed into mean words of 14.96 per text file. iii. A unique was assigned to each text file, and Non-Standard Kiswahili words, including abbreviations, monetary units, and numbers, were expanded into their full forms. Therefore, the text file has 3 parts: unique id, transcribed words, and normalized words. The text file was saved in CSV format. iv. A recording per text file was made by a native Kiswahili speaker using the audacity software in a least-noisy environment. v. The recording was saved as a WAVE file and assigned the same id as the text file.