Logo Lanfrica

Twi 8-Word Speech Segments

Domain:

natural language processing

Record type:

dataset
Creator:
gha
Host:
51139 speech-text pairs split from 30-min recordings. Source audio from ghananlpcommunity/ghana-female-twi-tts-full-length Full-file CTC forced alignment (MMS-300M) for word-level timestamps Words grouped into 8-word (8-gram) segments Leading/trailing silence trimmed with VAD (-40 dBFS threshold) Filtered: min 1.0s, max 15.0s Original sample rate preserved (24kHz) from datasets import load_dataset