A Bilingual Pashto Speech and Text Dataset for Low-Resource Language Research (BPSTD)

We introduce the Bilingual Pashtu Speech and Text Dataset (BPSTD) for low-resource language research. BPSTD is a bilingual speech-text dataset created to facilitate research in Pashto voice processing, speech recognition, and text-to-speech synthesis. The BPSTD consists of recordings from twenty native Pashto speakers, meticulously gathered under controlled conditions to guarantee clarity and consistency. It comprises 20K audio recordings, each accompanied by relevant textual information. The textual material includes Pashto script, its Romanized transcription, and an English translation. The dataset consists of two primary components: an audio folder containing recordings of all text, and a corresponding Excel file with three columns (Pashto text, Romanized Pashto, and English translation). We believe this resource will provide a strong foundation for advancing research in low-resource language modelling, speech synthesis, and cross-lingual applications for the Pashto language.