-
Notifications
You must be signed in to change notification settings - Fork 4
Expand file tree
/
Copy pathindex.html
More file actions
143 lines (108 loc) · 6.41 KB
/
index.html
File metadata and controls
143 lines (108 loc) · 6.41 KB
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
<!DOCTYPE html>
<html>
<head>
<meta charset='utf-8'>
<meta http-equiv="X-UA-Compatible" content="chrome=1">
<meta name="description" content="OpenDcd : An Open Source WFST based Speech Recognition Decoder">
<link rel="stylesheet" type="text/css" media="screen" href="stylesheets/stylesheet.css">
<title>OpenDcd</title>
</head>
<body>
<!-- HEADER -->
<div id="header_wrap" class="outer">
<header class="inner">
<a id="forkme_banner" href="https://github.com/opendcd">View on GitHub</a>
<h1 id="project_title">OpenDcd</h1>
<h2 id="project_tagline">An Open Source WFST based Speech Recognition Decoder</h2>
</header>
</div>
<!-- MAIN CONTENT -->
<div id="main_content_wrap" class="outer">
<section id="main_content" class="inner">
<h1>
<a id="opendcd---an-open-source-wfst-based-speech-recognition-decoder" class="anchor" href="#opendcd---an-open-source-wfst-based-speech-recognition-decoder" aria-hidden="true"><span class="octicon octicon-link"></span></a>OpenDcd - An Open Source WFST based Speech Recognition Decoder</h1>
<p>OpenDcd a lightweight and portable WFST based speech decoding toolkit written in C++. OpenDcd provides a set of tools for decoding, cascade construction and hypothesis post- processing. The focus of the toolkit is to provide a foundation for research into new decoding methods that can be deployed. Through the use of C++ templates the core decoder can be configured and extended in many ways. For example selecting different on-the-fly composition or lattice generation strategies. The core engine has detailed profiling, logging and analysis methods that make it highly for deployement in production systems. The toolkit makes used of <a href="http://openfst.org/">OpenFst</a> for representing and manipulating the models. It is distributed as an open source project with an Apache Licence. </p>
<p>For more information see the <a href="https://github.com/opendcd/opendcd.github.io/wiki">main documentation site</a>, and the tutorial for installing the OpenDcd and decoding using the Librispeech corpus and models from kaldi-asr.og.</p>
<h2>
<a id="quick-installation-guide" class="anchor" href="#quick-installation-guide" aria-hidden="true"><span class="octicon octicon-link"></span></a>Quick Installation Guide</h2>
<div class="highlight highlight-bash"><pre> git clone https://github.com/opendcd/opendcd.git</pre></div>
<div class="highlight highlight-bash"><pre> <span class="pl-s3">cd</span> src/bin
make</pre></div>
<h2>
<a id="kaldi-conversion-quick-start" class="anchor" href="#kaldi-conversion-quick-start" aria-hidden="true"><span class="octicon octicon-link"></span></a>Kaldi Conversion Quick Start</h2>
<p>For Kaldi model converion and decoding a working Kaldi installation and
set of acoustic and language models and features from generated from a Kaldi egs/s5
script are required. The following example is based on the output of Kaldi WSJ training run.</p>
<p>Graph construction, the scripts directory contains
The Kaldi language directory, we re-use the existing Kaldi lexicon and LM.</p>
<div class="highlight highlight-bash"><pre> <span class="pl-s3">cd</span> <span class="pl-vo">$OPENDCD</span>/scripts
<span class="pl-s">export</span> KALDI_ROOT=/home/opendcd/tools/kaldi-trunk
./makeclevel.sh \
<span class="pl-vo">$KALDI_ROOT</span>/egs/wsj/s5/data/lang_test_bg_5k \
<span class="pl-vo">$KALDI_ROOT</span>/egs/wsj/s5/exp/tri2a \
<span class="pl-vo">$KALDI_ROOT</span>/egs/wsj/s5/exp/ocd_tri2a \
<span class="pl-vo">$KALDI_ROOT</span></pre></div>
<p>See egs directory contains example script for showing how to convert a Kaldi WSJ setup</p>
<h2>
<a id="brief-overview" class="anchor" href="#brief-overview" aria-hidden="true"><span class="octicon octicon-link"></span></a>Brief Overview</h2>
<p>The first release includes the following features:</p>
<ul>
<li>Standalone lightweight decoder core</li>
<li>Kaldi file format compatible or ptionally build against Kaldi</li>
<li>Post-processing tools</li>
<li>OpenFst and Kaldi Interop Tools</li>
</ul>
<p>Decoder:</p>
<ul>
<li>Customizable transition model for custom user and transition models</li>
<li>Direct decoding on different weight semiring</li>
<li>On-the-fly decoding using lookahead composition</li>
<li>Lattice generation </li>
<li>Switchable STL implementations. Use different implementations such EASTL or RDESTL, or mix optimized containers such as Google sparse hash.</li>
<li>Powerful registration mechanism for adding user defined acoustic models and or lattice generation strategies</li>
</ul>
<p>Cascde construction:</p>
<ul>
<li>Script to efficiently build and convert models from a Kaldi lang directory</li>
</ul>
<p>Results post-processing:</p>
<ul>
<li>
<code>farfilter</code> Apply the command to every in FST in the FAR archive</li>
<li>
<code>latticetofar</code> Convert Kaldi Table to OpenFst FAR archive</li>
<li>
<code>fartolattice</code> Convert an OpenFst FAR archive to Kaldi Table</li>
</ul>
<p>Kaldi Interoperability:</p>
<ul>
<li>Write results in Kaldi <em>Lattice</em> table format</li>
<li>
<a href="https://github.com/opendcd/opendcd.github.io/wiki/Kaldi-Interoperability">More information</a> on optionally building against Kaldi </li>
<li>Convert Kaldi tree to optimized decoding structure</li>
</ul>
<p>More Information</p>
<ul>
<li>A <a href="https://github.com/opendcd/opendcd.github.io/wiki/EC2-Installation-Walkthrough">getting start guide</a> for running OpenDcd on Ec2 using the Librispeech models</li>
<li>Ongoing introdutory slides can be found <a href="https://dl.dropboxusercontent.com/u/321851/opendcd.pdf">here</a>. These are updated infrequently. </li>
</ul>
</section>
</div>
<!-- FOOTER -->
<div id="footer_wrap" class="outer">
<footer class="inner">
<p>Published with <a href="http://pages.github.com">GitHub Pages</a></p>
</footer>
</div>
<script type="text/javascript">
var gaJsHost = (("https:" == document.location.protocol) ? "https://ssl." : "http://www.");
document.write(unescape("%3Cscript src='" + gaJsHost + "google-analytics.com/ga.js' type='text/javascript'%3E%3C/script%3E"));
</script>
<script type="text/javascript">
try {
var pageTracker = _gat._getTracker("UA-53767021-1");
pageTracker._trackPageview();
} catch(err) {}
</script>
</body>
</html>