Speech interfaces for Pakistani languages.
Recognition, synthesis and voice access to digital information.
The challenge
Text-led services can exclude people who prefer spoken interaction or cannot comfortably read the language of an interface. Local-language speech tools need to connect recognition and synthesis with an accessible way to ask for and receive information.
Automatic speech recognition and text-to-speech connect spoken interaction to application information. Urdu, Punjabi, Pashto and Balochi form the language scope, with API access and a WhatsApp chatbot providing routes for using the models.
Contribution and outputs
The work combines speech recognition and synthesis for Urdu, Punjabi, Pashto and Balochi with API and WhatsApp interfaces.
- Speech recognition and synthesis models
- Urdu, Punjabi, Pashto and Balochi language scope
- API access for application integration
- A WhatsApp chatbot interface
Related work
An Arabic interface for specialised investment assistants.
AI Systems
Right-to-left product engineering, persistent conversations and separate knowledge contexts.
AI and risk systems across the customs lifecycle.
AI Systems
Four connected modules, from goods descriptions to post-clearance review.