I have recently completed a dissertation on creating topic-specific language models from English Wikipedia to improve recognition of specialist vocabulary for ASR of lectures, using Sphinx4 to evaluate the results against the reference HUB4 open source language model. It may be of interest to anyone working in the area of language model adaptation.
I'd like to extend my thanks to all the Sphinx4 contributors and developers, and especially Nickolay for his always-generous help in answering questions about Sphinx4 and speech recognition generally.
Regards
Stephen
If you would like to refer to this comment somewhere else in this project, copy and paste the following link:
Hi all,
I have recently completed a dissertation on creating topic-specific language models from English Wikipedia to improve recognition of specialist vocabulary for ASR of lectures, using Sphinx4 to evaluate the results against the reference HUB4 open source language model. It may be of interest to anyone working in the area of language model adaptation.
The abstract is here:
http://trulymadlywordly.blogspot.com/2012/12/improving-searchability-of.html
and the full text:
http://pubs.cs.uct.ac.za/archive/00000846/01/MPhil-Dissertation-StephenMarquard.pdf
I'd like to extend my thanks to all the Sphinx4 contributors and developers, and especially Nickolay for his always-generous help in answering questions about Sphinx4 and speech recognition generally.
Regards
Stephen
Great research, Stephen, thanks for sharing it!
Congratulations with graduation!