Commit 58b2d951 authored by Karel Vesely's avatar Karel Vesely
Browse files

sync with trunk



git-svn-id: https://svn.code.sf.net/p/kaldi/code/sandbox/karel@753 5e6a8d80-dfce-4ca6-a32a-6e07a63d50c8
parent 412161a3
......@@ -40,7 +40,7 @@ for($pdf = 0; $pdf < $numpdfs; $pdf++) {
print " <WEIGHTS> FV $nm [ ";
for($n = 0; $n < $nm; $n++) { print "$weight{$pdf,$n} "; }
print "]\n";
print " <MEANS_INVVARS> FM $nm $dim [\n";
print " <MEANS_INVVARS> [\n";
for($n = 0; $n < $nm; $n++) {
for($d = 0; $d < $dim; $d++) {
$val = $mean{$pdf,$n,$d} / $var{$pdf,$n,$d};
......@@ -49,7 +49,7 @@ for($pdf = 0; $pdf < $numpdfs; $pdf++) {
print "\n";
}
print " ]\n";
print " <INV_VARS> FM $nm $dim [\n";
print " <INV_VARS> [\n";
for($n = 0; $n < $nm; $n++) {
for($d = 0; $d < $dim; $d++) {
$val = 1.0 / $var{$pdf,$n,$d};
......
This diff is collapsed.
This diff is collapsed.
PAPER=kaldi_asru
all: $(PAPER).pdf
cp $(PAPER).pdf ~/desktop/2011_asru_kaldi.pdf
ifeq ($(shell uname), Darwin)
$(PAPER).pdf: $(PAPER).tex $(PAPER).bbl
pdflatex $(PAPER)
open $(PAPER).pdf
else
$(PAPER).pdf: $(PAPER).tex $(PAPER).bbl
pdflatex $(PAPER)
endif
$(PAPER).bbl: $(PAPER).tex refs.bib
pdflatex $(PAPER)
bibtex $(PAPER)
pdflatex $(PAPER)
clean:
-rm $(PAPER).aux $(PAPER).blg $(PAPER).bbl $(PAPER).log *~
.PHONY: clean
This diff is collapsed.
@conference{rwth,
Author = {David Rybach and Christian Gollan and Georg Heigold and Bj\"{o}rn Hoffmeister and Jonas L\"{o}\"{o}f and Ralf Schl\"{u}ter and Hermann Ney},
Booktitle = {INTERSPEECH},
Pages = {2111-2114},
Title = "{The RWTH Aachen University Open Source Speech Recognition System}",
Year = {2009}}
@book{htkbook,
author = {S. Young and G. Evermann and M. Gales and T. Hain and D. Kershaw and X. Liu and G. Moore and J. Odell and D. Ollason and D. Povey and V. Valtchev and P. Woodland},
title = "The HTK Book (for version 3.4)",
publisher = {Cambridge University Engineering Department},
year = {2009}}
@conference{julius,
Author = {Akinobu Lee and Tatsuya Kawahara and Kiyohiro Shikano},
Booktitle = {EUROSPEECH},
Pages = {1691-1694},
Title = {Julius -- an Open Source Real-Time Large Vocabulary Recognition Engine},
Year = {2001}}
@conference{attila,
Author = {Hagen Soltau and George Saon and Brian Kingsbury},
Booktitle = {IEEE Workshop on Spoken Language Technology (SLT)},
Pages = {97-102},
Title = "{The IBM Attila speech recognition toolkit}",
Year = {2010}}
@techreport{sphinx,
Author = {Willie Walker and Paul Lamere and Philip Kwok and Bhiksha Raj and Rita Singh and Evandro Gouvea and Peter Wolf and Joe Woelfel},
Institution = {Sun Microsystems Inc.},
Number = {SML1 TR2004-0811},
Title = {Sphinx-4: {A} Flexible Open Source Framework for Speech Recognition},
Type = {Technical Report},
Year = {2004}}
@article{sgmm_csl,
Author = {D. Povey and L. Burget and others},
Journal = {Computer Speech \& Language},
Title = {{The subspace Gaussian mixture model---A structured model for speech recognition}},
Volume = {25},
Number = {2},
Pages = {404-439},
Month = {April},
Year = {2011}}
@conference{goel_lex,
Author = {Nagendra Goel and Samuel Thomas and others},
Booktitle = {Proc. IEEE ICASSP},
Pages = {5094-5097},
Title = {Approaches to automatic lexicon learning with limited training examples},
Year = {2010}}
@conference{asru_et,
Author = {Daniel Povey and Geoffrey Zweig and Alex Acero},
Booktitle = {IEEE ASRU},
Title = {{The Exponential Transform as a generic substitute for VTLN}},
Year = {2011}}
@INPROCEEDINGS {openfst,
author = {C. Allauzen and M. Riley and J. Schalkwyk and W. Skut and M. Mohri },
title = "{OpenFst}: a general and efficient weighted finite-state transducer library",
booktitle = "Proc. CIAA",
year = 2007
}
@Book {golub,
author = "G. H. Golub and C. F. {Van Loan}",
title = "{Matrix Computations}",
edition = 3,
publisher = "{The Johns Hopkins University Press}",
year = 1996
}
@article {blas,
author = "L. S. Blackford and J. Demmel and J. Dongarra and I. Duff and S. Hammarling and G. Henry and M. Heroux and L. Kaufman and A. Lumsdaine and A. Petitet and R. Pozo and K. Remington and R. C. Whaley",
title = {An updated set of basic linear algebra subprograms ({BLAS})},
journal = {ACM Trans. Math. Softw.},
volume = {28},
issue = {2},
month = {June},
year = {2002},
pages = {135--151},
publisher = {ACM},
}
@UNPUBLISHED{atlas,
AUTHOR = "{ATLAS homepage}",
NOTE = "{http://math-atlas.sourceforge.net/}"
}
@ARTICLE{atlas2,
AUTHOR = "R. Clint Whaley and Antoine Petitet and Jack J. Dongarra",
TITLE = "Automated Empirical Optimization of Software and the
{ATLAS} Project",
JOURNAL = "Parallel Computing",
VOLUME = "27",
NUMBER = "1--2",
PAGES = "3--35",
YEAR = 2001}
@BOOK{lapack,
AUTHOR = {Anderson, E. and Bai, Z. and Bischof, C. and
Blackford, S. and Demmel, J. and Dongarra, J. and
Du Croz, J. and Greenbaum, A. and Hammarling, S. and
McKenney, A. and Sorensen, D.},
TITLE = {{LAPACK} Users' Guide},
EDITION = {Third},
PUBLISHER = {Society for Industrial and Applied Mathematics},
YEAR = {1999}}
@UNPUBLISHED{jama,
AUTHOR = "{JAMA homepage}",
NOTE = "{http://math.nist.gov/javanumerics/jama/}"
}
@TECHREPORT{ubmdoc,
author = {D. Povey},
title = "{A Tutorial-style Introduction to Subspace Gaussian Mixture Models for Speech Recognition}",
institution = "{Microsoft Research}",
year = 2009,
number = "MSR-TR-2009-111"
}
@inproceedings{stolcke2002srilm,
title={SRILM-an extensible language modeling toolkit},
author={Stolcke, A.},
booktitle={Proceedings of the international conference on spoken language processing},
volume={2},
pages={901--904},
year={2002},
}
@conference{irstlm,
Author = {Marcello Federico and Nicola Bertoldi and Mauro Cettolo},
Booktitle = {INTERSPEECH},
Pages = {1618-1621},
Title = "{IRSTLM}: An Open Source Toolkit for Handling Large Scale Language Models",
Year = {2008}}
@conference{Ghoshal:ICASSP10,
Author = {A. Ghoshal and D. Povey and others},
Booktitle = {Proc. IEEE ICASSP},
Pages = {4310-4313},
Title = {{A Novel Estimation of Feature-space MLLR for Full-covarinace Models}},
Year = {2010}}
@article{Gales:CSL98,
Author = {M. J. F. Gales},
Journal = {Computer Speech and Language},
Month = {April},
Number = {2},
Pages = {75-98},
Title = {Maximum likelihood linear transformations for {HMM}-based speech recognition},
Volume = {12},
Year = {1998}}
@conference{Price:ICASSP88,
Author = {P. Price and W. M. Fisher and J. Bernstein and D. S. Pallett},
Booktitle = {Proc. IEEE ICASSP},
Pages = {651-654},
Title = {The DARPA 1000-word resource management database for continuous speech recognition},
Volume = {1},
Year = {1988}}
@conference{Povey:ICASSP99,
Author = {D. Povey and P. C. Woodland},
Booktitle = {Proc. IEEE ICASSP},
Pages = {333-336},
Title = {Frame discrimination training for {HMMs} for large vocabulary speech recognition},
Volume = {1},
Year = {1999}}
@article{stc,
Author = {Mark J. F. Gales},
Journal = {IEEE Trans. Speech and Audio Proc.},
Month = {May},
Number = {3},
Pages = {272-281},
Title = {Semi-Tied Covariance Matrices for Hidden {Markov} Models},
Volume = {7},
Year = {1999}}
@conference{mllt,
Author = {Ramesh Gopinath},
Booktitle = {Proc. IEEE ICASSP},
Pages = {661-664},
Title = {Maximum likelihood modeling with {Gaussian} distributions for classification},
Volume = {2},
Year = {1998}}
@article{gales_linxform,
Author = {Mark J. F. Gales},
Journal = {Computer Speech and Language},
Month = {April},
Number = {2},
Pages = {75-98},
Title = {Maximum likelihood linear transformations for {HMM}-based speech recognition},
Volume = {12},
Year = {1998}}
@techreport{regtree,
Author = {Mark J. F. Gales},
Institution = {Cambridge University Engineering Department},
Month = {August},
Number = {CUED/F-INFENG/TR.263},
Title = {The Generation and Use of Regression Class Trees for {MLLR} Adaptation},
Type = {Technical Report},
Year = {1996}}
@article{hlda,
Author = {Nagendra Kumar and Andreas G. Andreou},
Journal = {Speech Communication},
Month = {December},
Number = {4},
Pages = {283-297},
Title = {Heteroscedastic discriminant analysis and reduced rank {HMMs} for improved speech recognition},
Volume = {26},
Year = {1998}}
@Article{wfst,
author = {Mehryar Mohri and Fernando Pereira and Michael Riley},
title = "Weighted Finite-State Transducers in Speech Recognition",
journal = "Computer Speech and Language",
year = {2002},
pages = {69--88},
volume = {20},
number = {1}
}
@INPROCEEDINGS {clustering,
author = "S. J. Young and J. J. Odell and P. C. Woodland",
title = {Tree-Based State Tying for High Accuracy Acoustic Modelling},
booktitle = "{Proc. 1994 ARPA Human Language Technology Workshop}",
year = 1994,
pages = "304--312"
}
@article{Reichl:ITSAP00,
Author = {Wolfgang Reichl and Wu Chou},
Journal = {IEEE Transactions on Speech and Audio Processing},
Month = {September},
Number = {5},
Pages = {555-566},
Title = {Robust decision tree state tying for continuous speech recognition},
Volume = {8},
Year = {2000}}
@conference{Woodland:ICASSP94,
Author = {P. C. Woodland and J. J. Odell and V. Valtchev and S. J. Young},
Booktitle = {Proc. IEEE ICASSP},
Pages = {II/125-II/128},
Title = {Large vocabulary continuous speech recognition using {HTK}},
Volume = {2},
Year = {1994}}
@Book {rico_book,
author = "Henrique S. Malvar",
title = "{Signal Processing with Lapped Transforms}",
publisher = {Artech House, Inc.},
year = 1992
}
@article{mllr,
Author = {C. J. Leggetter and P. C. Woodland},
Journal = {Computer Speech and Language},
Number = {2},
Pages = {171-185},
Title = {Maximum likelihood linear regression for speaker adaptation of continuous density hidden {Markov} models},
Volume = {9},
Year = {1995}}
@conference{lvtln,
Author = {D. Y. Kim and S. Umesh and M. J. F. Gales and T. Hain and P. C. Woodland},
Booktitle = {Proc. ICSLP},
Pages = {1953-1956},
Title = {Using {VTLN} for Broadcast News Transcription},
Year = {2004}}
all: kaldi-poster.pdf code-samples.ps script-toplevel.ps script-lowlevel.ps
kaldi-poster.pdf: kaldi-poster.tex beamerthemekaldi1.sty code-samples.pdf
pdflatex kaldi-poster
code-samples.pdf: code-samples.tex
pdflatex code-samples.tex
code-samples.ps: code-samples.pdf # only needed so we can use ghostview to find bounding box.
pdf2ps code-samples.pdf code-samples.ps
script-toplevel.pdf: script-toplevel.tex
pdflatex script-toplevel.tex
script-toplevel.ps: script-toplevel.pdf # only needed so we can use ghostview to find bounding box.
pdf2ps script-toplevel.pdf script-toplevel.ps
script-lowlevel.pdf: script-lowlevel.tex
pdflatex script-lowlevel.tex
script-lowlevel.ps: script-lowlevel.pdf # only needed so we can use ghostview to find bounding box.
pdf2ps script-lowlevel.pdf script-lowlevel.ps
#latex code-samples.tex
# dvips code-samples.dvi -o code-samples.eps
# mv code-samples.eps tmp.eps
# cat tmp.eps | awk '/BoundingBox/{ print "%%BoundingBox: 128 517 486 725"; next; } {print;}' \
> code-samples.eps
# ps2pdf -dEPSCrop code-samples.eps code-samples.pdf
This currently just contains a template I modified from one of Aachen University's.
It needs very updated LaTeX packages to compile-- try to find a machine where you can compile.
[dan]
% Copyright 2007 by
% Philippe Dreuw <dreuw@cs.rwth-aachen.de> and
% Thomas Deselaers <deselaers@cs.rwth-aachen.de>
%
% This file may be distributed and/or modified
%
% 1. under the LaTeX Project Public License and/or
% 2. under the GNU Public License.
%
%
% ChangeLog:
%
% 1.10 - imrpoved PackageWarning and PackageInfo outputs
% 1.09 - bugfixed the list indention problem, can be overwritten by user with the \setdefaultleftmargin command.
% - added 'printer' option for external printer definition files
% 1.08 - supports external printer definition files
% - added grid mode option
% - renamed beamer specific variables
% - added font size normalization, so that scale=1.0 is the default font size for all DIN-A(n) sizes
% 1.07 - bugfixed custom size handling, portrait or landscape settings are ignored now
% 1.06 - added the type1cm package for scalable math fonts
% 1.05 - added version check for xkeyval package
% 1.04 - added custom size handling
% 1.03 - improved predefined size handling
% 1.02 - minor bugfixes
% 1.01 - bugfixed size handling
% 1.00 - first beamerposter release
%
\def\beamerposter@name{beamerposter}
\def\beamerposter@version{1.10}
\def\beamerposter@date{2009/09/02}
\def\beamerposter@msg{latex-beamer poster extension}
\def\beamerposter@releaseinfo{\beamerposter@date \space v.\beamerposter@version \space \beamerposter@msg}
\def\beamerposter@Error#1#2{\PackageError{\beamerposter@name}{#1}{#2}}
\def\beamerposter@Warning#1{\PackageWarning{\beamerposter@name}{#1}}
\def\beamerposter@Info#1{\PackageInfo{\beamerposter@name}{#1}}
\NeedsTeXFormat{LaTeX2e}
\ProvidesPackage{\beamerposter@name}[\beamerposter@releaseinfo]
\RequirePackage{xkeyval}[2006/11/18]
\RequirePackage{type1cm} %% get it from ftp://cam.ctan.org/tex-archive/macros/latex/contrib/type1cm.zip
\RequirePackage{fp}
\RequirePackage{paralist}
\newif\if@beamerposter@portrait
\newif\if@beamerposter@customsize
\newif\if@beamerposter@extern
\newif\if@beamerposter@debug
\DeclareOptionX{size}[a0]{
\beamerposter@Info{checking size input, please wait.}
\XKV@cc*+[\val\nr]{#1}{a0b,a0,a1,a2,a3,a4,custom}{%
\beamerposter@Info{the input \val\ \nr\ was correct, we proceed.}
\ifcase\nr\relax
%a0b
\edef\paperwidthValue{119}
\edef\paperheightValue{88}
\edef\textwidthValue{116}
\edef\textheightValue{88}
\edef\fontscale{1.0}
\or
%a0
\edef\paperwidthValue{118.82}
\edef\paperheightValue{83.96}
\edef\textwidthValue{117.82}
\edef\textheightValue{82.96}
\edef\fontscale{1.0}
\or
%a1
\edef\paperwidthValue{83.96}
\edef\paperheightValue{59.4}
\edef\textwidthValue{82.96}
\edef\textheightValue{58.4}
\edef\fontscale{0.70710678} % fontscale=(1/sqrt(2))^1
\or
%a2
\edef\paperwidthValue{59.4}
\edef\paperheightValue{41.98}
\edef\textwidthValue{58.4}
\edef\textheightValue{40.98}
\edef\fontscale{0.5} % fontscale=(1/sqrt(2))^2
\or
%a3
\edef\paperwidthValue{41.98}
\edef\paperheightValue{29.7}
\edef\textwidthValue{40.98}
\edef\textheightValue{28.7}
\edef\fontscale{0.35355339} % fontscale=(1/sqrt(2))^3
\or
%a4
\edef\paperwidthValue{29.7}
\edef\paperheightValue{21.0}
\edef\textwidthValue{28.7}
\edef\textheightValue{20.0}
\edef\fontscale{0.25} % fontscale=(1/sqrt(2))^4
\or
\@beamerposter@customsizetrue
\fi
}{%
\beamerposter@Warning{the input \val\ was incorrect and was ignored.}
}%
\beamerposter@Info{finished size input check.}
}
\DeclareOptionX{orientation}[portrait]{
\beamerposter@Info{checking orientation input, please wait.}
\XKV@cc*+[\val\nr]{#1}{portrait,landscape}{%
\beamerposter@Info{the input \val\ \nr\ was correct, we proceed.}
\ifcase\nr\relax
\@beamerposter@portraittrue
\or
\@beamerposter@portraitfalse
\fi
}{%
\beamerposter@Warning{the input \val\ was incorrect and was ignored.}
}%
\beamerposter@Info{finished orientation check.}
}
\DeclareOptionX{scale}[1.0]{\edef\myfontscale{#1}\beamerposter@Info{myfontscale=\myfontscale}}
\DeclareOptionX{width}{\edef\customwidth{#1}\beamerposter@Info{custom poster width=\customwidth}}
\DeclareOptionX{height}{\edef\customheight{#1}\beamerposter@Info{custom poster height=\customheight}}
\DeclareOptionX{debug}{\beamerposter@Info{enabled debug mode}\@beamerposter@debugtrue}
\DeclareOptionX{grid}{\beamerposter@Info{enabling grid}\beamertemplategridbackground[1cm]}% Display a grid to help align images
\DeclareOptionX{printer}{\edef\printerToUse{#1}\@beamerposter@externtrue}
\DeclareOptionX*{\beamerposter@Warning{Unknown option ignored: \CurrentOption}}
%\DeclareOptionX*{\PassOptionsToClass{\CurrentOption}{beamer}}
\ExecuteOptionsX{size=a0,scale=1.0}
\ProcessOptionsX\relax
%% enable debug mode for fp package (mainly for upn functions)
\if@beamerposter@debug
\FPdebugtrue
\fi
%% normalize scale depending on poster size
\FPupn{\myfontscale}{myfontscale fontscale * 2 round}
%% swap sizes for portrait orientation
\if@beamerposter@portrait
\newdimen\tmp
\setlength{\tmp}{\paperwidthValue cm}
\setlength{\paperwidth}{\paperheightValue cm}
\setlength{\paperheight}{\tmp}
\setlength{\tmp}{\textwidthValue cm}
\setlength{\textwidth}{\textheightValue cm}
\setlength{\textheight}{\tmp}
\else
\setlength{\paperwidth}{\paperwidthValue cm}
\setlength{\paperheight}{\paperheightValue cm}
\setlength{\textwidth}{\textwidthValue cm}
\setlength{\textheight}{\textheightValue cm}
\fi
%% overwrite dimensions if custom size
\if@beamerposter@customsize
\setlength{\paperwidth}{\customwidth cm}
\setlength{\paperheight}{\customheight cm}
\FPupn{\resulttextwidth}{1 customwidth -}
\FPupn{\resulttextheight}{1 customheight -}
\setlength{\textwidth}{\resulttextwidth cm}
\setlength{\textheight}{\resulttextheight cm}
\fi
%% Setting proper dimensions for a DIN A0 printer
\setlength{\headheight}{0 cm}
\setlength{\headsep}{0 cm}
\setlength{\topmargin}{-12.7 mm} % -1in +1.47cm
\setlength{\oddsidemargin}{-25.4 mm} % -1in +0.4cm
%% For the page layout
\beamerposter@Info{paperwidth=\the\paperwidth, paperheight=\the\paperheight}
\beamerposter@Info{textwidth=\the\textwidth, textwidth=\the\textheight}
\beamerposter@Info{fontscale=\fontscale, myfontscale=\myfontscale}
\geometry{
paperwidth=\the\paperwidth,
paperheight=\the\paperheight,
hmargin=1cm,%
vmargin=0cm,%
head=0.5cm, %
headsep=0pt,%
foot=0.5cm %
}
%%
%% printer definition file example 'rwth-glossy-uv.df':
%% ----
%% \typeout{beamerposter: RWTH RZ printer, glossy paper, UV resistant, maximum settings}
%% \edef\printermaxwidthValue{108} %unit should be cm
%% \newlength{\printermaxwidth}
%% \setlength{\printermaxwidth}{\printermaxwidthValue cm}
%% ----
%%
%% now load printer definitions from file (e.g. \usepackage[printer=rwth-glossy-uv.df]{beamerposter}")
\if@beamerposter@extern
\InputIfFileExists{\printerToUse}{\typeout{load \printerToUse}}{\beamerposter@Error{\printerToUse\space not found!}{Please use another printer definition file}}
\ifx\printermaxwidthValue\@empty%
\relax%
\else%
%% @todo: calc poster ratios, scale to maximum printer size, and check if they still fit
%% this here is still very basic ...
\FPsub{\resultmaxwidth}{\printermaxwidthValue}{\paperwidthValue}
\FPifneg{\resultmaxwidth}
\beamerposter@Error{%
resultmaxwidth=\resultmaxwidth \MessageBreak
printermaxwidthValue=\printermaxwidthValue \MessageBreak
paperwidthValue=\paperwidthValue \MessageBreak
The specified poster dimensions might not be printable using
'\printerToUse'.
}{%
Try to reduce your poster dimension and have a look at the aspect ratio.
}
\else
\relax
\fi
\fi
%\fi