25 years of speech technology innovation

Topics covered
Popular Clips
Episode Highlights
Cambodian Project
The Cambodian Stories Project is a remarkable initiative aimed at preserving the oral histories of Cambodians, particularly those affected by the Khmer Rouge atrocities. explains how his team at Cobalt Speech and Language collaborated with Brigham Young University (BYU) to develop a Cambodian speech recognition system. This project not only preserves cultural heritage but also advances speech technology for lesser-resourced languages 1.
We have here a situation where lesser resource languages, where we have now thousands of hours of high quality recordings with good transcriptions for a big chunk of them, I thought, those are the ingredients you need to develop a speech recognizer.
---
The collaboration involves students learning about speech technology while contributing to a meaningful cause, highlighting the potential for similar projects in other languages 2.
  Â
Data Challenges
Acquiring data for building speech technology for less-represented languages presents significant challenges. notes that access to appropriate data is a major obstacle in developing speech recognition systems, especially for languages with fewer speakers 3.
It's hard to get, for these lesser represented languages, it's hard to get sufficient data that's labeled appropriately to be able to build the kind of technology that we already are starting to enjoy.
---
This challenge underscores the need for innovative solutions to bring the benefits of speech technology to all language communities, not just the major ones.
Related Episodes


Speech recognition to say it just right
Answers 383 questions

Accelerating ML innovation at MLCommons
Answers 383 questions

The first real-time voice assistant
Answers 383 questions

AI adoption in large, well-established companies
Answers 383 questions

Next-gen voice assistants
Answers 383 questions

2019's AI top 5
Answers 383 questions

Accidentally building SOTA AI
Answers 383 questions

AI adoption in the enterprise
Answers 383 questions

Speech tech and Common Voice at Mozilla
Answers 383 questions

From symbols to AI pair programmers 💻
Answers 383 questions

Generative models: exploration to deployment
Answers 383 questions

AI predictions for 2024
Answers 383 questions

Success (and failure) in prompting
Answers 383 questions

IBM's AI for detecting neurological state
Answers 383 questions

Mapping the world
Answers 383 questions
