Skip to content
GitLab
Explore
Sign in
Primary navigation
Search or go to…
Project
K
kaldi-commonvoice
Manage
Activity
Members
Labels
Plan
Issues
Issue boards
Milestones
Wiki
Code
Merge requests
Repository
Branches
Commits
Tags
Repository graph
Compare revisions
Snippets
Build
Pipelines
Jobs
Pipeline schedules
Artifacts
Deploy
Releases
Container Registry
Model registry
Operate
Environments
Monitor
Incidents
Analyze
Value stream analytics
Contributor analytics
CI/CD analytics
Repository analytics
Model experiments
Help
Help
Support
GitLab documentation
Compare GitLab plans
Community forum
Contribute to GitLab
Provide feedback
Keyboard shortcuts
?
Snippets
Groups
Projects
Show more breadcrumbs
Simon Will
kaldi-commonvoice
Commits
68c5c058
Commit
68c5c058
authored
9 years ago
by
Daniel Povey
Committed by
Dan Povey
9 years ago
Browse files
Options
Downloads
Patches
Plain Diff
Some minor changes and renaming of files; starting documentation
parent
c4d67a74
No related branches found
Branches containing commit
No related tags found
No related merge requests found
Changes
3
Hide whitespace changes
Inline
Side-by-side
Showing
3 changed files
src/ctc/cctc-supervision.cc
+0
-0
0 additions, 0 deletions
src/ctc/cctc-supervision.cc
src/ctc/cctc-supervision.h
+7
-5
7 additions, 5 deletions
src/ctc/cctc-supervision.h
src/doc/ctc.dox
+51
-0
51 additions, 0 deletions
src/doc/ctc.dox
with
58 additions
and
5 deletions
src/ctc/ctc-supervision.cc
→
src/ctc/
c
ctc-supervision.cc
+
0
−
0
View file @
68c5c058
File moved
This diff is collapsed.
Click to expand it.
src/ctc/ctc-supervision.h
→
src/ctc/
c
ctc-supervision.h
+
7
−
5
View file @
68c5c058
// lat/ctc-supervision.h
// lat/
c
ctc-supervision.h
// Copyright 2015 Johns Hopkins University (Author: Daniel Povey)
...
...
@@ -34,7 +34,9 @@
namespace
kaldi
{
// CTC means Connectionist Temporal Classification, see the paper by Graves et
// al.
// al. CCTC means context-dependent CTC, which
//
//
// What we are implementing here is some things relating to computation of the
// CTC objective function. The normal information required to compute the
...
...
@@ -201,9 +203,9 @@ void AddBlanksToProtoSupervision(const CtcSupervisionOptions &options,
- Exactly one copy of that phone symbol.
- Respectively (before and after) each real phone symbol mentioned above,
optional repeats of the blank symbol (symbol 0). The time information
for these
instances of the blank symbol will be the same as the phone
symbol we
generated it from.
optional repeats of the blank symbol (symbol 0). The time information
for these
instances of the blank symbol will be the same as the phone
symbol we
generated it from.
This fst will have a state 0 which is both initial and final. Let's
suppose we have a phone-in-context a. first an arc from state 0 to a new
...
...
This diff is collapsed.
Click to expand it.
src/doc/ctc.dox
0 → 100644
+
51
−
0
View file @
68c5c058
// doc/ctc.dox
// Copyright 2015 Johns Hopkins University (author: Daniel Povey)
// See ../../COPYING for clarification regarding multiple authors
//
// Licensed under the Apache License, Version 2.0 (the "License");
// you may not use this file except in compliance with the License.
// You may obtain a copy of the License at
// http://www.apache.org/licenses/LICENSE-2.0
// THIS CODE IS PROVIDED *AS IS* BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY
// KIND, EITHER EXPRESS OR IMPLIED, INCLUDING WITHOUT LIMITATION ANY IMPLIED
// WARRANTIES OR CONDITIONS OF TITLE, FITNESS FOR A PARTICULAR PURPOSE,
// MERCHANTABLITY OR NON-INFRINGEMENT.
// See the Apache 2 License for the specific language governing permissions and
// limitations under the License.
namespace kaldi {
/**
\page ctc Connectionist Temporal Classifiction
\section Introduction
This page describes the support for Connectionist Temporal Classifiction (CTC)
in Kaldi. At the current time CTC is only supported in the nnet3 setup, but
the non-neural-net parts of the CTC code are designed to be independent of the
neural net implementation so porting it to other setups such as nnet1 and nnet2
should be possible.
The original CTC paper is "Connectionist temporal classification: labelling
unsegmented sequence data with recurrent neural networks" by Alex Graves et al.
The core of CTC is somewhat indepenent of of the exact network topology
Originally the idea of CTC was to have the network output words directly, but
more recent work by Google, e.g.
"Fast and Accurate Recurrent Neural Network Acoustic Models for Speech
Recognition" by Hasim Sak, Andrew Senior et al.
Todo: finish this intro.
*/
}
This diff is collapsed.
Click to expand it.
Preview
0%
Loading
Try again
or
attach a new file
.
Cancel
You are about to add
0
people
to the discussion. Proceed with caution.
Finish editing this message first!
Save comment
Cancel
Please
register
or
sign in
to comment