Bilingual Telugu-Gondi Corpus

Swecha

About this app

Swecha Gondi Corpus is a field data-collection tool for building a
bilingual Telugu-Gondi parallel speech corpus.

For each prompt, a volunteer records a short spoken answer in Telugu
and the same answer in Gondi. The app runs on-device speech
recognition immediately after each recording, so the speaker can
review and correct the transcript before moving on.

Once a session is complete, volunteers validate their recordings —
playing each clip back, editing the transcript if needed, and
accepting or rejecting the pair. Accepted recordings are uploaded to
corpus.swecha.org, where the Telugu and Gondi halves of each sentence
are linked together as a matched pair rather than stored as two
unrelated clips.

Features:
• Guided recording sessions with configurable size
• On-device speech recognition for instant transcription
• Playback and inline transcript correction before upload
• Automatic linking of Telugu/Gondi recording pairs
• Session progress tracking (in progress → ready to validate → uploaded)

Swecha is an open-source community building free and open datasets,
tools, and software for Indian languages.