Javaabu demonstrates new Dhivehi Speech Transcriber with automatic speaker diarization

Malé, Maldives, September 12, 2023 Javaabu has released an early preview of its experimental Dhivehi Speech Transcriber, showcasing a major advancement in automated audio processing for the Dhivehi language. The latest prototype can now automatically segment audio by individual speakers, a capability known as speaker diarization.

In a demo shared by the team, the system successfully identified and separated contributions from different speakers in a Majlis recording. It also demonstrated the ability to track the same speaker across multiple segments of a conversation, even when their speech appears at different points in the audio.

This feature is part of Javaabu’s broader research effort to develop accurate, Dhivehi-first speech technologies that can support transcription, archiving, media workflows, and accessibility tools. The team noted that the diarization system is still experimental but represents an important step toward more advanced Dhivehi speech-to-text solutions.