Commit bb9d15ee authored by Dan Povey's avatar Dan Povey
Browse files

Final edits to paper.

git-svn-id: https://svn.code.sf.net/p/kaldi/code/sandbox/discrim@536 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8
parent 08af6717
...@@ -92,7 +92,7 @@ there are obvious similarities. ...@@ -92,7 +92,7 @@ there are obvious similarities.
In the experiments we report here we have not been able to match the performance of In the experiments we report here we have not been able to match the performance of
a standard diagonal GMM based system using a traditional semi-continuous system; a standard diagonal GMM based system using a traditional semi-continuous system;
however, using a two-level tree and multiple codebooks similar to~\cite{prasad2004t2b}, however, using a two-level tree and multiple codebooks similar to~\cite{prasad2004t2b},
we were able to get better performance than a traditional system (although not in one setup we were able to get better performance than a traditional system (although not
as good as SGMMs). For both traditional and multiple-codebook configurations, as good as SGMMs). For both traditional and multiple-codebook configurations,
we saw better results with full-covariance codebook Gaussians than with diagonal. we saw better results with full-covariance codebook Gaussians than with diagonal.
We investigated smoothing the statistics for weight estimation using the decision We investigated smoothing the statistics for weight estimation using the decision
...@@ -542,7 +542,8 @@ and the other using multiple codebooks based on a two-level phonetic decision tr ...@@ -542,7 +542,8 @@ and the other using multiple codebooks based on a two-level phonetic decision tr
similar to~\cite{prasad2004t2b}. similar to~\cite{prasad2004t2b}.
Using the single-codebook system we were not able to obtain competitive Using the single-codebook system we were not able to obtain competitive
results, but using a multiple-codebook system results, but using a multiple-codebook system
we were able to get better results than a traditional system. Interestingly, we were able to get better results than a traditional system, for one
database (RM). Interestingly,
for both the single-codebook or multiple-codebook systems, we saw better for both the single-codebook or multiple-codebook systems, we saw better
results with full rather than diagonal covariances. results with full rather than diagonal covariances.
...@@ -567,7 +568,8 @@ the one we used here. ...@@ -567,7 +568,8 @@ the one we used here.
%\label{sec:ref} %\label{sec:ref}
% Korbinian: We might need that ;) % Korbinian: We might need that ;)
\footnotesize %\footnotesize
\bibliographystyle{IEEEbib} \bibliographystyle{IEEEbib}
\bibliography{refs-eig,refs} \bibliography{refs-eig,refs}
......
Markdown is supported
0% or .
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment