Add CHiME4 recipe #746
Merged
Add CHiME4 recipe #746
Conversation
This file contains bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
|
Will update results in README.md in the later days. |
robin1001
reviewed
Nov 11, 2021
wenet/dataset/dataset.py
Outdated
| @@ -139,7 +139,7 @@ def Dataset(data_type, data_list_file, symbol_table, conf, | |||
| else: | |||
| dataset = Processor(dataset, processor.parse_raw) | |||
|
|
|||
| dataset = Processor(dataset, processor.tokenize, symbol_table, bpe_model) | |||
| dataset = Processor(dataset, processor.tokenize, symbol_table, bpe_model, conf.get('char', False)) | |||
robin1001
requested changes
Nov 11, 2021
wenet/dataset/processor.py
Outdated
| @@ -258,7 +258,7 @@ def compute_fbank(data, | |||
| yield dict(key=sample['key'], label=sample['label'], feat=mat) | |||
|
|
|||
|
|
|||
| def tokenize(data, symbol_table, bpe_model=None): | |||
| def tokenize(data, symbol_table, bpe_model=None, char=False): | |||
There was a problem hiding this comment.
tokenize use char as the default model unit.
You mean your model unit is seperated with white space? it is more like phoneme, such as timit. It's better we take another name.
There was a problem hiding this comment.
I'd like to use split_with_space instead
robin1001
reviewed
Nov 12, 2021
examples/chime4/s0/README.md
Outdated
| | decoding mode | dt05_real_1ch | dt05_simu_1ch | et05_real_1ch | et05_simu_1ch | | ||
| |:---------------:|:-------------:|:-------------:|:-------------:|:-------------:| | ||
| | ctc_beam_search | 19.06% | 21.17% | 28.39% | 29.16% | | ||
| | att_rescoring | 17.92% | 20.22% | 27.40% | 28.25% | |
There was a problem hiding this comment.
ctc_beam_search -> ctc_prefix_beam_search
att_rescoring -> attention_rescoring
rename so it is consistent with other recipes.
robin1001
approved these changes
Nov 19, 2021
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Add CHiME4 recipe with character unit. Training data combines WSJ0+1 & CHiME4 real & simulation utterances. Current best WER with CTC prefix beam search are 21.17%/19.06%/28.41%/29.16% on dt05 real/simu and et05 real/simu, respectively.