Building Inclusive Speech Recognition for All in Kenya

In today’s digital world, voice technology has become a powerful tool for communication, accessibility, and convenience. From virtual assistants to voice-enabled applications, speaking to devices is now second nature for many. But what happens when voice recognition doesn’t work for everyone? For individuals with speech impairments, this technology often remains out of reach, creating a digital divide that limits participation and inclusion.

The GDI Hub’s Center for Digital Language Inclusion (CDLI) is working to change that. Through its Automatic Speech Recognition (ASR) project in Kenya, CDLI is developing speech recognition models tailored for people living with diverse speech patterns, including those with impaired speech. The goal is simple yet transformative: to ensure that voice technology is available and inclusive for everyone not just those whose speech fits conventional models.

Starting with Data

Every innovation begins with data, and this project is no exception. In Nairobi, the team has been meeting with individuals with varying degrees of speech impairment some mild, others severe to record voice samples. These recordings will form the foundation for training ASR models that can understand Swahili speakers with speech differences.

This work is critical because, historically, digital applications have offered little utility for persons with speech impairments. Even when they can use these tools, language barriers and recognition inaccuracies often limit their effectiveness. By creating specialized models, the project aims to break these barriers and empower individuals to communicate confidently and effectively.

This matters because communication is a fundamental human right, not a privilege. For people with speech impairments, being understood can mean more than convenience it can mean dignity, independence, and inclusion. Enhanced ASR technology will:

• Boost confidence: Allow individuals to articulate needs and interact with society without fear of being misunderstood.
• Enable cooperation: Foster more effective conversations between people with speech impairments and those around them.
• Promote equality: Ensure that technological advancements benefit everyone, regardless of age, location, or gender.