Active and Semi-Supervised Learning for Automatic Speech Recognition
Lead Research Organisation:
University of Cambridge
Department Name: Engineering
Abstract
The project aims to develop machine learning techniques that allow deep learning based speech recognition systems to be effectively trained on large amounts of un-labelled or partially labelled data. Additionally, it aims to develop techniques for systems to actively chose an approximately optimal subset of the un-labelled data to be transcribed by humans for further improvement of the system.
This project fits well into the EPSRC research areas "Artificial intelligence technologies" and "Speech technology".
This project fits well into the EPSRC research areas "Artificial intelligence technologies" and "Speech technology".
Organisations
People |
ORCID iD |
Philip Woodland (Primary Supervisor) | |
Florian Kreyssig (Student) |