Hi Solvers,
We’ve recently published a reference solution for the Children’s Speech Recognition Challenge: Word Track.
In this in-depth blog post and companion repo, we walk through how to fine-tune NVIDIA’s Parakeet ASR model using NeMo on the competition data, resulting in a 0.2370 WER on the public leaderboard.
In the tutorial, we:
- Demonstrate how to load and explore the data.
- Provide a basic framework for building a model.
- Demonstrate how to package your work correctly for submission.
Whether you’re just getting started or looking to benchmark your approach, this should give you a strong foundation to build on. Read the full post here:
https://drivendata.co/blog/child-asr-word-benchmark
Happy modeling!