Automatic Speech Recognition
Welsh
kaldi
vosk
welsh
cymraeg
speech-recognition
DewiBrynJones commited on
Commit
c1adedb
·
verified ·
1 Parent(s): 467a245

Release 2610 from techiaith/kaldi-cy-2610@main

Browse files
Files changed (4) hide show
  1. LICENSE +202 -0
  2. NOTICE +126 -0
  3. README.md +88 -23
  4. model_cy.tar.gz +2 -2
LICENSE ADDED
@@ -0,0 +1,202 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+
2
+ Apache License
3
+ Version 2.0, January 2004
4
+ http://www.apache.org/licenses/
5
+
6
+ TERMS AND CONDITIONS FOR USE, REPRODUCTION, AND DISTRIBUTION
7
+
8
+ 1. Definitions.
9
+
10
+ "License" shall mean the terms and conditions for use, reproduction,
11
+ and distribution as defined by Sections 1 through 9 of this document.
12
+
13
+ "Licensor" shall mean the copyright owner or entity authorized by
14
+ the copyright owner that is granting the License.
15
+
16
+ "Legal Entity" shall mean the union of the acting entity and all
17
+ other entities that control, are controlled by, or are under common
18
+ control with that entity. For the purposes of this definition,
19
+ "control" means (i) the power, direct or indirect, to cause the
20
+ direction or management of such entity, whether by contract or
21
+ otherwise, or (ii) ownership of fifty percent (50%) or more of the
22
+ outstanding shares, or (iii) beneficial ownership of such entity.
23
+
24
+ "You" (or "Your") shall mean an individual or Legal Entity
25
+ exercising permissions granted by this License.
26
+
27
+ "Source" form shall mean the preferred form for making modifications,
28
+ including but not limited to software source code, documentation
29
+ source, and configuration files.
30
+
31
+ "Object" form shall mean any form resulting from mechanical
32
+ transformation or translation of a Source form, including but
33
+ not limited to compiled object code, generated documentation,
34
+ and conversions to other media types.
35
+
36
+ "Work" shall mean the work of authorship, whether in Source or
37
+ Object form, made available under the License, as indicated by a
38
+ copyright notice that is included in or attached to the work
39
+ (an example is provided in the Appendix below).
40
+
41
+ "Derivative Works" shall mean any work, whether in Source or Object
42
+ form, that is based on (or derived from) the Work and for which the
43
+ editorial revisions, annotations, elaborations, or other modifications
44
+ represent, as a whole, an original work of authorship. For the purposes
45
+ of this License, Derivative Works shall not include works that remain
46
+ separable from, or merely link (or bind by name) to the interfaces of,
47
+ the Work and Derivative Works thereof.
48
+
49
+ "Contribution" shall mean any work of authorship, including
50
+ the original version of the Work and any modifications or additions
51
+ to that Work or Derivative Works thereof, that is intentionally
52
+ submitted to Licensor for inclusion in the Work by the copyright owner
53
+ or by an individual or Legal Entity authorized to submit on behalf of
54
+ the copyright owner. For the purposes of this definition, "submitted"
55
+ means any form of electronic, verbal, or written communication sent
56
+ to the Licensor or its representatives, including but not limited to
57
+ communication on electronic mailing lists, source code control systems,
58
+ and issue tracking systems that are managed by, or on behalf of, the
59
+ Licensor for the purpose of discussing and improving the Work, but
60
+ excluding communication that is conspicuously marked or otherwise
61
+ designated in writing by the copyright owner as "Not a Contribution."
62
+
63
+ "Contributor" shall mean Licensor and any individual or Legal Entity
64
+ on behalf of whom a Contribution has been received by Licensor and
65
+ subsequently incorporated within the Work.
66
+
67
+ 2. Grant of Copyright License. Subject to the terms and conditions of
68
+ this License, each Contributor hereby grants to You a perpetual,
69
+ worldwide, non-exclusive, no-charge, royalty-free, irrevocable
70
+ copyright license to reproduce, prepare Derivative Works of,
71
+ publicly display, publicly perform, sublicense, and distribute the
72
+ Work and such Derivative Works in Source or Object form.
73
+
74
+ 3. Grant of Patent License. Subject to the terms and conditions of
75
+ this License, each Contributor hereby grants to You a perpetual,
76
+ worldwide, non-exclusive, no-charge, royalty-free, irrevocable
77
+ (except as stated in this section) patent license to make, have made,
78
+ use, offer to sell, sell, import, and otherwise transfer the Work,
79
+ where such license applies only to those patent claims licensable
80
+ by such Contributor that are necessarily infringed by their
81
+ Contribution(s) alone or by combination of their Contribution(s)
82
+ with the Work to which such Contribution(s) was submitted. If You
83
+ institute patent litigation against any entity (including a
84
+ cross-claim or counterclaim in a lawsuit) alleging that the Work
85
+ or a Contribution incorporated within the Work constitutes direct
86
+ or contributory patent infringement, then any patent licenses
87
+ granted to You under this License for that Work shall terminate
88
+ as of the date such litigation is filed.
89
+
90
+ 4. Redistribution. You may reproduce and distribute copies of the
91
+ Work or Derivative Works thereof in any medium, with or without
92
+ modifications, and in Source or Object form, provided that You
93
+ meet the following conditions:
94
+
95
+ (a) You must give any other recipients of the Work or
96
+ Derivative Works a copy of this License; and
97
+
98
+ (b) You must cause any modified files to carry prominent notices
99
+ stating that You changed the files; and
100
+
101
+ (c) You must retain, in the Source form of any Derivative Works
102
+ that You distribute, all copyright, patent, trademark, and
103
+ attribution notices from the Source form of the Work,
104
+ excluding those notices that do not pertain to any part of
105
+ the Derivative Works; and
106
+
107
+ (d) If the Work includes a "NOTICE" text file as part of its
108
+ distribution, then any Derivative Works that You distribute must
109
+ include a readable copy of the attribution notices contained
110
+ within such NOTICE file, excluding those notices that do not
111
+ pertain to any part of the Derivative Works, in at least one
112
+ of the following places: within a NOTICE text file distributed
113
+ as part of the Derivative Works; within the Source form or
114
+ documentation, if provided along with the Derivative Works; or,
115
+ within a display generated by the Derivative Works, if and
116
+ wherever such third-party notices normally appear. The contents
117
+ of the NOTICE file are for informational purposes only and
118
+ do not modify the License. You may add Your own attribution
119
+ notices within Derivative Works that You distribute, alongside
120
+ or as an addendum to the NOTICE text from the Work, provided
121
+ that such additional attribution notices cannot be construed
122
+ as modifying the License.
123
+
124
+ You may add Your own copyright statement to Your modifications and
125
+ may provide additional or different license terms and conditions
126
+ for use, reproduction, or distribution of Your modifications, or
127
+ for any such Derivative Works as a whole, provided Your use,
128
+ reproduction, and distribution of the Work otherwise complies with
129
+ the conditions stated in this License.
130
+
131
+ 5. Submission of Contributions. Unless You explicitly state otherwise,
132
+ any Contribution intentionally submitted for inclusion in the Work
133
+ by You to the Licensor shall be under the terms and conditions of
134
+ this License, without any additional terms or conditions.
135
+ Notwithstanding the above, nothing herein shall supersede or modify
136
+ the terms of any separate license agreement you may have executed
137
+ with Licensor regarding such Contributions.
138
+
139
+ 6. Trademarks. This License does not grant permission to use the trade
140
+ names, trademarks, service marks, or product names of the Licensor,
141
+ except as required for reasonable and customary use in describing the
142
+ origin of the Work and reproducing the content of the NOTICE file.
143
+
144
+ 7. Disclaimer of Warranty. Unless required by applicable law or
145
+ agreed to in writing, Licensor provides the Work (and each
146
+ Contributor provides its Contributions) on an "AS IS" BASIS,
147
+ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or
148
+ implied, including, without limitation, any warranties or conditions
149
+ of TITLE, NON-INFRINGEMENT, MERCHANTABILITY, or FITNESS FOR A
150
+ PARTICULAR PURPOSE. You are solely responsible for determining the
151
+ appropriateness of using or redistributing the Work and assume any
152
+ risks associated with Your exercise of permissions under this License.
153
+
154
+ 8. Limitation of Liability. In no event and under no legal theory,
155
+ whether in tort (including negligence), contract, or otherwise,
156
+ unless required by applicable law (such as deliberate and grossly
157
+ negligent acts) or agreed to in writing, shall any Contributor be
158
+ liable to You for damages, including any direct, indirect, special,
159
+ incidental, or consequential damages of any character arising as a
160
+ result of this License or out of the use or inability to use the
161
+ Work (including but not limited to damages for loss of goodwill,
162
+ work stoppage, computer failure or malfunction, or any and all
163
+ other commercial damages or losses), even if such Contributor
164
+ has been advised of the possibility of such damages.
165
+
166
+ 9. Accepting Warranty or Additional Liability. While redistributing
167
+ the Work or Derivative Works thereof, You may choose to offer,
168
+ and charge a fee for, acceptance of support, warranty, indemnity,
169
+ or other liability obligations and/or rights consistent with this
170
+ License. However, in accepting such obligations, You may act only
171
+ on Your own behalf and on Your sole responsibility, not on behalf
172
+ of any other Contributor, and only if You agree to indemnify,
173
+ defend, and hold each Contributor harmless for any liability
174
+ incurred by, or claims asserted against, such Contributor by reason
175
+ of your accepting any such warranty or additional liability.
176
+
177
+ END OF TERMS AND CONDITIONS
178
+
179
+ APPENDIX: How to apply the Apache License to your work.
180
+
181
+ To apply the Apache License to your work, attach the following
182
+ boilerplate notice, with the fields enclosed by brackets "[]"
183
+ replaced with your own identifying information. (Don't include
184
+ the brackets!) The text should be enclosed in the appropriate
185
+ comment syntax for the file format. We also recommend that a
186
+ file or class name and description of purpose be included on the
187
+ same "printed page" as the copyright notice for easier
188
+ identification within third-party archives.
189
+
190
+ Copyright [yyyy] [name of copyright owner]
191
+
192
+ Licensed under the Apache License, Version 2.0 (the "License");
193
+ you may not use this file except in compliance with the License.
194
+ You may obtain a copy of the License at
195
+
196
+ http://www.apache.org/licenses/LICENSE-2.0
197
+
198
+ Unless required by applicable law or agreed to in writing, software
199
+ distributed under the License is distributed on an "AS IS" BASIS,
200
+ WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
201
+ See the License for the specific language governing permissions and
202
+ limitations under the License.
NOTICE ADDED
@@ -0,0 +1,126 @@
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
1
+ kaldi-cy 2610, released October 2026
2
+ Copyright 2026 Bangor University
3
+
4
+ Developed by the Language Technologies Unit (Uned Technolegau Iaith), Bangor University.
5
+
6
+ Licensed under the Apache License, Version 2.0 (the "License"). You may obtain a copy of the
7
+ License at http://www.apache.org/licenses/LICENSE-2.0
8
+
9
+ NAMES
10
+ This licence grants no permission to use the names "Bangor University", "Prifysgol Bangor",
11
+ "Language Technologies Unit", "Uned Technolegau Iaith", "techiaith", the University's logos, or
12
+ the names of its products or staff, except to describe the origin of the model as this NOTICE
13
+ does.
14
+
15
+ WHAT THIS MODEL IS
16
+ This is a research model, trained on the data and evaluated by the methods described in its
17
+ model card. It is not a certified product and is not a substitute for a qualified translator,
18
+ transcriber, clinician or other professional.
19
+
20
+ AUTOMATIC PROCESSING MAKES MISTAKES
21
+ Output from this model, whether a translation, a transcript, generated text or speech, or any
22
+ other result, can be wrong while appearing fluent and confident. Known failure modes include
23
+ omitted or inserted words, incorrect names, numbers, dates and negations, and reduced accuracy
24
+ on dialects, noisy audio or subjects outside the training data.
25
+
26
+ Prifysgol Bangor University therefore advises that output is treated as provisional and is
27
+ checked and corrected by a competent person before it is relied on as a record, used in any
28
+ decision affecting a person, or shared or published. Where the model is used interactively or in
29
+ real time (for example live captions, dictation, suggestions in an editor or read-aloud), users
30
+ should be told that what they see or hear is automatic and may be wrong. The University asks
31
+ anyone who builds on or redistributes this model to pass this warning on to their own users.
32
+
33
+ NOT FIT FOR
34
+ The model has not been evaluated for, and should not be used in, medical, legal, financial,
35
+ safety-critical or emergency settings, or in any decision affecting a person's rights, health,
36
+ liberty or livelihood, unless a qualified person reviews every output.
37
+
38
+ USE IN WALES: EQUAL TREATMENT OF WELSH AND ENGLISH
39
+ This model was developed to help organisations provide services in Welsh that are as good as
40
+ those they provide in English, not to replace Welsh-language provision. Under section 1 of the
41
+ Welsh Language (Wales) Measure 2011 (2011 nawm 1), Welsh has official status in Wales, and that
42
+ status is given legal effect by, among other things, the treatment of Welsh no less favourably
43
+ than English (s. 1(2)(b)) and by standards of conduct relating to the use of Welsh (s. 1(3)(e)).
44
+ These standards are made and enforced under Part 4 of the Measure. Bodies still operating a
45
+ Welsh language scheme under Part II of the Welsh Language Act 1993 (c. 38) must give effect to
46
+ the principle in section 5 of that Act that, in the conduct of public business and the
47
+ administration of justice in Wales, Welsh and English should be treated on a basis of equality.
48
+ Prifysgol Bangor University asks anyone deploying this model in Wales to consider whether the
49
+ intended use meets those requirements. Offering automated Welsh where English is served by
50
+ people, publishing machine translation into Welsh without the review given to English text, or
51
+ otherwise using this technology to replace rather than support Welsh-language provision is
52
+ contrary to the purpose for which the model was released. Which duties apply depends on the body
53
+ and the regulations or scheme that cover it, and responsibility for compliance rests with the
54
+ body deploying the model. See the Welsh Language Commissioner's regulatory policy statement on
55
+ artificial intelligence and the Welsh language, Technology, Websites and Software: Welsh
56
+ Language Considerations, and the advice document Bilingual drafting, translation and using Welsh
57
+ face to face, all linked from the model card. The legislation is at
58
+ legislation.gov.uk/mwa/2011/1 and legislation.gov.uk/ukpga/1993/38.
59
+
60
+ FURTHER INFORMATION AND ERROR REPORTS
61
+ Training data, evaluation results and known limitations: https://huggingface.co/techiaith/kaldi-cy
62
+ Report errors or concerns to: techiaith@bangor.ac.uk
63
+
64
+ PRONUNCIATION DICTIONARY
65
+ The decoding graph (HCLG.fst) and word list (words.txt) in this bundle were built with
66
+ pronunciations from Geiriadur Ynganu Bangor, the Bangor Pronouncing Dictionary,
67
+ https://github.com/techiaith/geiriadur-ynganu-bangor, redistributed here under its own
68
+ licence, which requires the following to be kept with any binary redistribution:
69
+
70
+ Copyright (C) 2025 Prifysgol Bangor University. All rights reserved.
71
+
72
+ Redistribution and use in source and binary forms, with or without modification, are
73
+ permitted provided that the following conditions are met:
74
+
75
+ 1. Redistributions of source code must retain the above copyright notice, this list of
76
+ conditions and the following disclaimer. The contents of this file are deemed to be
77
+ source code.
78
+
79
+ 2. Redistributions in binary form must reproduce the above copyright notice, this list
80
+ of conditions and the following disclaimer in the documentation and/or other
81
+ materials provided with the distribution.
82
+
83
+ THIS SOFTWARE IS PROVIDED BY PRIFYSGOL BANGOR UNIVERSITY ``AS IS'' AND ANY EXPRESSED OR
84
+ IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
85
+ MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL
86
+ PRIFYSGOL BANGOR UNIVERSITY NOR ITS EMPLOYEES BE LIABLE FOR ANY DIRECT, INDIRECT,
87
+ INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO,
88
+ PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
89
+ INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT
90
+ LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE
91
+ OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
92
+
93
+ BUILT WITH
94
+ Trained with Kaldi (Apache License 2.0) and pocolm (Apache License 2.0), and packaged
95
+ for Vosk (Apache License 2.0), following the vosk-cymraeg recipe by Gweltaz
96
+ Duval-Guennoc, Preben Vangberg and Sasha Wanasky (MIT License, Cymru-Breizh Agile Cymru
97
+ Project, https://github.com/Cymru-Breizh-Agile-Cymru-Project/vosk-cymraeg). None of
98
+ their code is included in this bundle.
99
+
100
+ TRAINING DATA ATTRIBUTIONS
101
+ The Apache License applies to the model weights, code and documentation. It does not license the
102
+ data the model was trained on, which is not distributed with the model and remains under the
103
+ terms of its original providers, acknowledged here as those terms require.
104
+
105
+ Dedicated to the public domain under CC0 1.0 Universal
106
+ (https://creativecommons.org/publicdomain/zero/1.0/). CC0 asks for no attribution; the sources
107
+ are listed so that the training data can be traced to its datasheet.
108
+ Banc Trawsgrifiadau Bangor (CC0 1.0), collected by Language Technologies Unit, Bangor
109
+ University. Collection process, consent terms and contact:
110
+ https://huggingface.co/datasets/techiaith/banc-trawsgrifiadau-bangor
111
+ Corpws CLlLC / WLGA (CC0 1.0), collected by Language Technologies Unit, Bangor University.
112
+ Collection process, consent terms and contact:
113
+ https://huggingface.co/datasets/techiaith/corpws-clllc-wlga
114
+ Mozilla Common Voice 23.0, Welsh (CC0 1.0), collected by Mozilla Foundation and the volunteers
115
+ who contributed their voices. Collection process, consent terms and contact:
116
+ https://commonvoice.mozilla.org/cy/datasets. Obtained from Mozilla under their terms for
117
+ Common Voice data and held in a private working copy; it is not redistributed with or by this
118
+ model.
119
+ Lleisiau ARFOR (CC0 1.0), collected by Cymen, with the Language Technologies Unit, Bangor
120
+ University. Collection process, consent terms and contact:
121
+ https://huggingface.co/datasets/cymen-arfor/lleisiau-arfor
122
+ Mozilla Common Voice, Welsh, re-segmented with voice activity detection (CC0 1.0), collected by
123
+ Mozilla Foundation and the volunteers who contributed their voices. Collection process,
124
+ consent terms and contact: https://commonvoice.mozilla.org/cy/datasets. Derived from Common
125
+ Voice data obtained from Mozilla under their terms and held in a private working copy; it is
126
+ not redistributed with or by this model.
README.md CHANGED
@@ -1,5 +1,5 @@
1
  ---
2
- license: cc-by-nc-3.0
3
  language:
4
  - cy
5
  pipeline_tag: automatic-speech-recognition
@@ -13,10 +13,9 @@ datasets:
13
  - techiaith/banc-trawsgrifiadau-bangor
14
  - techiaith/corpws-clllc-wlga
15
  - cymen-arfor/lleisiau-arfor
16
- - techiaith/commonvoice_vad_cy
17
  ---
18
 
19
- # Kaldi / Vosk Welsh Speech Recognition — 2607
20
 
21
  A Kaldi chain (LF-MMI) acoustic model with WFST decoding, packaged for
22
  [Vosk](https://alphacephei.com/vosk/). It runs in real time on CPU, which makes it
@@ -24,32 +23,55 @@ suited to on-device and streaming use where an end-to-end model is too costly.
24
 
25
  Welsh transcription only. Output carries no punctuation or casing.
26
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
27
  ## Evaluation results
28
 
29
- WER / CER (lower is better) on the held-out `techiaith/asr-benchmark-cy-en-2606` test sets, scored
30
- per-utterance against the same references for every model compared.
 
 
31
 
32
- Mean across the Welsh test sets: **WER 47.25**, against **50.52** for the previous release (2503) — an improvement of **3.27** points.
33
 
34
- | Test set | WER | CER | WER (2503) |
35
  |----------|-----|-----|------------|
36
- | `cymen-arfor/lleisiau-arfor` (spontaneous, mixed dialects) | 55.99 | 28.05 | 54.45 |
37
- | `techiaith/banc-trawsgrifiadau-bangor` (spontaneous, mixed dialects) | 40.23 | 17.49 | 46.03 |
38
- | `techiaith/commonvoice-23-0-cy` (read) | 45.52 | 15.92 | 51.08 |
 
39
 
40
  ## Training data
41
 
42
  Welsh speech corpora used to train this model:
43
 
44
- | Dataset | Split | Revision | Access |
45
- |---------|-------|----------|--------|
46
- | [techiaith/banc-trawsgrifiadau-bangor](https://huggingface.co/datasets/techiaith/banc-trawsgrifiadau-bangor/tree/26.07) | `train` | `26.07` | public |
47
- | [techiaith/corpws-clllc-wlga](https://huggingface.co/datasets/techiaith/corpws-clllc-wlga/tree/5d00294c31c78b1d7937bb2c2bc6cc70bc18d410) | `clips` | `5d00294c31` | public |
48
- | `techiaith/commonvoice_23_0_cy` | `train+dev[50%:]` | `main` | private † |
49
- | [cymen-arfor/lleisiau-arfor](https://huggingface.co/datasets/cymen-arfor/lleisiau-arfor) | `train` | `main` | public |
50
- | [techiaith/commonvoice_vad_cy](https://huggingface.co/datasets/techiaith/commonvoice_vad_cy) | `train` | `main` | public |
51
 
52
- † **Common Voice data is not redistributed here.** `techiaith/commonvoice_23_0_cy` (Common Voice 23.0 Welsh) is held in a private repository, in line with the Mozilla Foundation's terms for obtaining and sharing Common Voice data. It is listed above for provenance and reproducibility, not as a download.
53
 
54
  The Welsh Common Voice corpus is freely available from Mozilla themselves: [Mozilla Common Voice — Welsh datasets](https://commonvoice.mozilla.org/cy/datasets). Releases are distributed through the Mozilla Data Collective, and the [`datacollective`](https://pypi.org/project/datacollective/) package offers a programmatic download. With that in hand the rest of this pipeline is reproducible — every other dataset in the table is openly available on the Hugging Face Hub. Our thanks to Mozilla and to the Welsh speakers who contributed their voices.
55
 
@@ -74,10 +96,10 @@ Welsh transcription audio only, as materialised for training:
74
  | Epochs | 5 |
75
  | Features | 40-dim MFCC (hires) + 100-dim online i-vectors |
76
  | SpecAugment | yes (spec-augment-layer in the xconfig) |
77
- | Language model | 3-gram, Witten-Bell discounting, trained on the training transcripts only |
78
  | Lexicon | Geiriadur Ynganu Bangor pronunciations, with grapheme-to-phoneme fallback |
79
  | Decoding vocabulary | 62,392 words |
80
- | Held-out dev WER | 30.91% (BTB 32.60%, Common Voice 24.85%) |
81
 
82
  ## Usage
83
 
@@ -117,15 +139,58 @@ returns `True`, as above. A single `AcceptWaveform` call followed by only
117
  `FinalResult()` discards everything recognised before the last detected
118
  end-of-utterance, which silently truncates anything longer than one utterance.
119
 
120
- ## Limitations
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
121
 
122
  - **Welsh only.** English audio is not a supported task.
123
  - **Closed vocabulary.** WFST decoding can only emit words present in the
124
  lexicon (62,392 words), which is built from the training transcripts. Names and
125
- loanwords outside it cannot be produced at all — a hard floor on achievable WER.
 
126
  - **No punctuation or casing.**
127
  - The language model is a 3-gram trained on the training transcripts only, so
128
- the model is weaker on domains and registers unlike its training data.
 
 
 
 
 
 
 
 
 
 
 
 
129
 
130
  ## Acknowledgements
131
 
 
1
  ---
2
+ license: apache-2.0
3
  language:
4
  - cy
5
  pipeline_tag: automatic-speech-recognition
 
13
  - techiaith/banc-trawsgrifiadau-bangor
14
  - techiaith/corpws-clllc-wlga
15
  - cymen-arfor/lleisiau-arfor
 
16
  ---
17
 
18
+ # Kaldi / Vosk Welsh Speech Recognition — 2610
19
 
20
  A Kaldi chain (LF-MMI) acoustic model with WFST decoding, packaged for
21
  [Vosk](https://alphacephei.com/vosk/). It runs in real time on CPU, which makes it
 
23
 
24
  Welsh transcription only. Output carries no punctuation or casing.
25
 
26
+ ## What changed in this release
27
+
28
+ **The language model is estimated with [pocolm](https://github.com/danpovey/pocolm)
29
+ instead of SRILM.** The training data, recipe, lexicon and acoustic model are those of
30
+ the 2607 release; only the 3-gram language model was re-estimated, and the decoding
31
+ graph rebuilt from it. This has two consequences.
32
+
33
+ - **Licence.** SRILM's Research Community License permits non-commercial research use
34
+ only, so every earlier Kaldi/Vosk release from the Unit (2503, 2607) was published
35
+ under CC-BY-NC. pocolm is Apache-2.0, as are Kaldi and Vosk, and the training speech
36
+ is all CC0, so nothing in this model's toolchain or data restricts its use. This is
37
+ the first Kaldi/Vosk model from the Unit released under the Apache License 2.0 (see
38
+ [Licence](#licence) below), and it may be used commercially.
39
+ - **Accuracy.** pocolm tunes its discounting against the held-out dev transcripts
40
+ instead of applying fixed Witten-Bell discounting, and keeps the n-grams that SRILM's
41
+ default count cut-offs discard. Held-out dev WER falls from 30.91% to 30.05%, and the
42
+ benchmark improves on every Welsh test set (table below). The larger language model
43
+ makes the decoding graph (`HCLG.fst`) three times the size, 242 MB against 82 MB, so
44
+ the bundle grows from 70 MB to 121 MB and decoding needs correspondingly more memory.
45
+
46
  ## Evaluation results
47
 
48
+ WER / CER (lower is better) on the held-out test sets, scored per utterance against the
49
+ same references for every model compared.
50
+
51
+ The references are the test sets' transcripts as published, with their capitalisation and punctuation intact; only transcription-convention markers (such as `<chwerthin>`) are removed before scoring. This model emits lowercase text without punctuation, so every reference word that carries a capital letter or attached punctuation is scored as an error. The figures are therefore conservative, and higher than a normalised comparison would give, but every model in the table is scored the same way, so the comparison between them holds.
52
 
53
+ Mean across the Welsh test sets: **WER 47.30**, against **48.20** for the previous release (2607-SRILM) — an improvement of **0.90** points.
54
 
55
+ | Test set | WER | CER | WER (2607-SRILM) |
56
  |----------|-----|-----|------------|
57
+ | `cymen-arfor/lleisiau-arfor` (spontaneous, mixed dialects) | 55.10 | 27.55 | 55.99 |
58
+ | `google/fleurs/cy-gb` | 49.89 | 20.25 | 51.06 |
59
+ | `techiaith/banc-trawsgrifiadau-bangor` (spontaneous, mixed dialects) | 39.69 | 17.25 | 40.23 |
60
+ | `techiaith/commonvoice-23-0-cy` (read) | 44.54 | 15.48 | 45.52 |
61
 
62
  ## Training data
63
 
64
  Welsh speech corpora used to train this model:
65
 
66
+ | Dataset | Split | Revision | Licence | Access | Datasheet |
67
+ |---------|-------|----------|---------|--------|-----------|
68
+ | [techiaith/banc-trawsgrifiadau-bangor](https://huggingface.co/datasets/techiaith/banc-trawsgrifiadau-bangor/tree/26.07) | `train` | `26.07` | [CC0 1.0](https://creativecommons.org/publicdomain/zero/1.0/) | public | [datasheet](https://huggingface.co/datasets/techiaith/banc-trawsgrifiadau-bangor) |
69
+ | [techiaith/corpws-clllc-wlga](https://huggingface.co/datasets/techiaith/corpws-clllc-wlga/tree/5d00294c31c78b1d7937bb2c2bc6cc70bc18d410) | `clips` | `5d00294c31` | [CC0 1.0](https://creativecommons.org/publicdomain/zero/1.0/) | public | [datasheet](https://huggingface.co/datasets/techiaith/corpws-clllc-wlga) |
70
+ | `techiaith/commonvoice_23_0_cy` | `train+dev[50%:]` | `main` | [CC0 1.0](https://creativecommons.org/publicdomain/zero/1.0/) | private &dagger; | [datasheet](https://commonvoice.mozilla.org/cy/datasets) |
71
+ | [cymen-arfor/lleisiau-arfor](https://huggingface.co/datasets/cymen-arfor/lleisiau-arfor) | `train` | `main` | [CC0 1.0](https://creativecommons.org/publicdomain/zero/1.0/) | public | [datasheet](https://huggingface.co/datasets/cymen-arfor/lleisiau-arfor) |
72
+ | `techiaith/commonvoice_vad_cy` | `train` | `main` | [CC0 1.0](https://creativecommons.org/publicdomain/zero/1.0/) | private &dagger; | [datasheet](https://commonvoice.mozilla.org/cy/datasets) |
73
 
74
+ &dagger; **Common Voice data is not redistributed here.** `techiaith/commonvoice_23_0_cy` (Common Voice 23.0 Welsh), `techiaith/commonvoice_vad_cy` (Common Voice Welsh, re-segmented with voice activity detection) are held in private repositories, in line with the Mozilla Foundation's terms for obtaining and sharing Common Voice data. They are listed above for provenance and reproducibility, not as a download.
75
 
76
  The Welsh Common Voice corpus is freely available from Mozilla themselves: [Mozilla Common Voice — Welsh datasets](https://commonvoice.mozilla.org/cy/datasets). Releases are distributed through the Mozilla Data Collective, and the [`datacollective`](https://pypi.org/project/datacollective/) package offers a programmatic download. With that in hand the rest of this pipeline is reproducible — every other dataset in the table is openly available on the Hugging Face Hub. Our thanks to Mozilla and to the Welsh speakers who contributed their voices.
77
 
 
96
  | Epochs | 5 |
97
  | Features | 40-dim MFCC (hires) + 100-dim online i-vectors |
98
  | SpecAugment | yes (spec-augment-layer in the xconfig) |
99
+ | Language model | 3-gram estimated with pocolm on the training transcripts only, discounting tuned on the dev transcripts (dev perplexity 292.8) |
100
  | Lexicon | Geiriadur Ynganu Bangor pronunciations, with grapheme-to-phoneme fallback |
101
  | Decoding vocabulary | 62,392 words |
102
+ | Held-out dev WER | 30.05% |
103
 
104
  ## Usage
105
 
 
139
  `FinalResult()` discards everything recognised before the last detected
140
  end-of-utterance, which silently truncates anything longer than one utterance.
141
 
142
+ ## Intended use and limitations
143
+
144
+ The model transcribes Welsh speech to text. It is a research model, not a certified product,
145
+ and is not a substitute for a qualified transcriber.
146
+
147
+ **Automatic transcription makes mistakes.** Output can be wrong while appearing fluent and
148
+ confident. Known failure modes include omitted, inserted or substituted words; incorrect names, numbers and place names; words outside its fixed vocabulary replaced by similar-sounding ones; and reduced accuracy on dialect, overlapping speakers, noisy or distant audio, and subjects outside the training data. Output carries no punctuation or capitalisation. Treat output as provisional and have it
149
+ checked and corrected by a competent person before it is relied on as a record, used in any
150
+ decision affecting a person, or shared or published. Where the model is used interactively or
151
+ in real time (live captions, dictation), tell users that what they see is an automatic
152
+ transcription and may be wrong.
153
+
154
+ **Not fit for** medical, legal, financial, safety-critical or emergency settings, or any
155
+ decision affecting a person's rights, health, liberty or livelihood, unless a qualified person
156
+ reviews every output.
157
+
158
+ **Use in Wales.** The model was developed to help organisations provide services in Welsh that
159
+ are as good as those they provide in English, not to replace Welsh-language provision. Under
160
+ section 1 of the Welsh Language (Wales) Measure 2011 (2011 nawm 1), Welsh has official status in
161
+ Wales, and that status is given legal effect by, among other things, the treatment of Welsh no
162
+ less favourably than English (s. 1(2)(b)) and by standards of conduct relating to the use of
163
+ Welsh (s. 1(3)(e)). These standards are made and enforced under Part 4 of the Measure. Bodies
164
+ still operating a Welsh language scheme under Part II of the Welsh Language Act 1993 (c. 38)
165
+ must give effect to the principle in section 5 of that Act that, in the conduct of public
166
+ business and the administration of justice in Wales, Welsh and English should be treated on a
167
+ basis of equality. Which duties apply depends on the body and the regulations or scheme that
168
+ cover it, and responsibility for compliance rests with the body deploying the model. See the Welsh Language Commissioner's [regulatory policy statement on artificial intelligence and the Welsh language](https://www.welshlanguagecommissioner.wales/regulation/promoting-compliance/artificial-intelligence-and-the-welsh-language-the-welsh-language-commissioners-regulatory-policy-statement), [Technology, Websites and Software: Welsh Language Considerations](https://www.welshlanguagecommissioner.wales/media/lxechbz0/technoleg-gwefannau-a-meddalwedd-technology-websites-and-software.pdf), and the advice document [Bilingual drafting, translation and using Welsh face to face](https://www.welshlanguagecommissioner.wales/support-resources/translation).
169
+ The legislation is at [legislation.gov.uk/mwa/2011/1](https://www.legislation.gov.uk/mwa/2011/1)
170
+ and [legislation.gov.uk/ukpga/1993/38](https://www.legislation.gov.uk/ukpga/1993/38).
171
+
172
+ Further limitations:
173
 
174
  - **Welsh only.** English audio is not a supported task.
175
  - **Closed vocabulary.** WFST decoding can only emit words present in the
176
  lexicon (62,392 words), which is built from the training transcripts. Names and
177
+ loanwords outside it cannot be produced at all and this represents a hard floor on
178
+ achievable WER.
179
  - **No punctuation or casing.**
180
  - The language model is a 3-gram trained on the training transcripts only, so
181
+ the model is weaker on domains and registers that differ from its training data.
182
+
183
+ ## Licence
184
+
185
+ Released under the [Apache License 2.0](LICENSE). The accompanying [NOTICE](NOTICE) file
186
+ carries the University's attribution, the warning that automatic transcription makes
187
+ mistakes, and the attributions required by the training-data and pronunciation-dictionary
188
+ licences. Section 4 of the licence requires anyone who redistributes the model or a
189
+ derivative to keep that NOTICE. Both files are also inside `model_cy.tar.gz`, in the `model/`
190
+ directory, so they travel with the unpacked model.
191
+
192
+ One-line attribution: *Includes kaldi-cy 2610, developed by the Language Technologies Unit,
193
+ Bangor University, and released under the Apache License 2.0.*
194
 
195
  ## Acknowledgements
196
 
model_cy.tar.gz CHANGED
@@ -1,3 +1,3 @@
1
  version https://git-lfs.github.com/spec/v1
2
- oid sha256:961ecedc17e2fbd01b720e916515bf33c03a59147eb46c1e8375e8c6425c58c1
3
- size 69963916
 
1
  version https://git-lfs.github.com/spec/v1
2
+ oid sha256:ddb8ea827262db10191f486a9a021687d95106b9e1c018aa45165ac5ab284919
3
+ size 120503544