Polishing the language of README.md
Browse files
README.md
CHANGED
@@ -71,13 +71,13 @@ Please refer to our [paper](https://www.isca-speech.org/archive/pdfs/interspeech
|
|
71 |
|
72 |
### Supported Tasks and Leaderboards
|
73 |
|
74 |
-
The MOCKS dataset can be used for Open-Vocabulary Keyword Spotting (OV-KWS) task. It supports two OV-KWS types:
|
75 |
-
- Query-by-Text, where keyword is provided by text and needs to be detected
|
76 |
-
- Query-by-Example, where keyword is provided with enrollment audio for detection
|
77 |
|
78 |
It also allows for:
|
79 |
-
- offline keyword detection, where test audio is
|
80 |
-
- online (streaming) keyword detection, where test audio
|
81 |
|
82 |
### Languages
|
83 |
|
@@ -90,7 +90,7 @@ The MOCKS incorporates 5 languages:
|
|
90 |
|
91 |
## Dataset Structure
|
92 |
|
93 |
-
The MOCKS testset is split by language, source dataset and OV-KWS type:
|
94 |
```
|
95 |
MOCKS
|
96 |
│
|
@@ -138,25 +138,25 @@ MOCKS
|
|
138 |
```
|
139 |
|
140 |
Each split is divided into:
|
141 |
-
- positive examples (`all.pair.positive.tsv`) - test examples with true
|
142 |
-
- similar examples (`all.pair.similar.tsv`) - test examples with similar phrases to keyword selected based on phonetic transcription distance,
|
143 |
-
- different examples (`all.pair.different.tsv`) - test examples with
|
144 |
|
145 |
All those files contain columns separated by tab:
|
146 |
- `keyword_path` - path to audio containing keyword phrase.
|
147 |
- `adversary_keyword_path` - path to test audio.
|
148 |
-
- `adversary_keyword_timestamp_start` - start time in seconds of phrase of interest for given keyword from `keyword_path`, field only available in **offline** split.
|
149 |
-
- `adversary_keyword_timestamp_end` - end time in seconds of phrase of interest for given keyword from `keyword_path`, field only available in **offline** split.
|
150 |
- `label` - whether the `adversary_keyword_path` contain keyword from `keyword_path` or not (1 - contains keyword, 0 - doesn't contain keyword).
|
151 |
|
152 |
-
Each split also contains subset of whole data with the same field
|
153 |
|
154 |
-
Also,
|
155 |
- `data_offline_transcription.tsv` - transcriptions for **offline** examples and `keyword_path` from **online** scenario,
|
156 |
-
- `data_online_transcription.tsv` - transcriptions for adversary, test examples from **online** scenario,
|
157 |
|
158 |
three columns are present within each file:
|
159 |
-
- `path_to_keyword`/`path_to_adversary_keyword` - path to audio file,
|
160 |
- `keyword_transcription`/`adversary_keyword_transcription` - audio transcription,
|
161 |
- `keyword_phonetic_transcription`/`adversary_keyword_phonetic_transcription` - audio phonetic transcription.
|
162 |
|
@@ -166,8 +166,10 @@ The dataset can be used by:
|
|
166 |
- downloading the archive and constructing all the test cases based on the provided `tsv` files,
|
167 |
- `datasets` package.
|
168 |
|
169 |
-
In the latter case the following should work:
|
170 |
-
```
|
|
|
|
|
171 |
|
172 |
The allowed values for `name` are:
|
173 |
- `en.LS-{clean,other}`,
|
@@ -218,13 +220,13 @@ All the test files are provided in 16 kHz, even though `{de,en,es,fr,it}.MCV` fi
|
|
218 |
|
219 |
The MOCKS testset was created from LibriSpeech and Mozilla Common Voice (MCV) datasets that are publicly available. To create it:
|
220 |
- a [MFA](https://mfa-models.readthedocs.io/en/latest/acoustic/index.html) with publicly available models was used to extract word-level alignments,
|
221 |
-
- an internally
|
222 |
|
223 |
The data is stored in a 16-bit, single-channel WAV format. 16kHz sampling rate is used for LibriSpeech based testset
|
224 |
and 48kHz sampling rate for MCV based testset.
|
225 |
|
226 |
-
The offline testset contains additional 0.1
|
227 |
-
The online version
|
228 |
|
229 |
The MOCKS testset is gender balanced.
|
230 |
|
|
|
71 |
|
72 |
### Supported Tasks and Leaderboards
|
73 |
|
74 |
+
The MOCKS dataset can be used for the Open-Vocabulary Keyword Spotting (OV-KWS) task. It supports two OV-KWS types:
|
75 |
+
- Query-by-Text, where the keyword is provided by text and needs to be detected in the audio stream.
|
76 |
+
- Query-by-Example, where the keyword is provided with enrollment audio for detection in the audio stream.
|
77 |
|
78 |
It also allows for:
|
79 |
+
- offline keyword detection, where test audio is trimmed to contain only keywords of interest.
|
80 |
+
- online (streaming) keyword detection, where test audio has past and future context besides keywords of interest.
|
81 |
|
82 |
### Languages
|
83 |
|
|
|
90 |
|
91 |
## Dataset Structure
|
92 |
|
93 |
+
The MOCKS testset is split by language, source dataset, and OV-KWS type:
|
94 |
```
|
95 |
MOCKS
|
96 |
│
|
|
|
138 |
```
|
139 |
|
140 |
Each split is divided into:
|
141 |
+
- positive examples (`all.pair.positive.tsv`) - test examples with true keywords, 5000-8000 keywords in each subset,
|
142 |
+
- similar examples (`all.pair.similar.tsv`) - test examples with similar phrases to the keyword selected based on phonetic transcription distance,
|
143 |
+
- different examples (`all.pair.different.tsv`) - test examples with completely different phrases.
|
144 |
|
145 |
All those files contain columns separated by tab:
|
146 |
- `keyword_path` - path to audio containing keyword phrase.
|
147 |
- `adversary_keyword_path` - path to test audio.
|
148 |
+
- `adversary_keyword_timestamp_start` - start time in seconds of phrase of interest for a given keyword from `keyword_path`, the field only available in **offline** split.
|
149 |
+
- `adversary_keyword_timestamp_end` - end time in seconds of phrase of interest for a given keyword from `keyword_path`, the field only available in **offline** split.
|
150 |
- `label` - whether the `adversary_keyword_path` contain keyword from `keyword_path` or not (1 - contains keyword, 0 - doesn't contain keyword).
|
151 |
|
152 |
+
Each split also contains a subset of whole data with the same field structure to allow faster evaluation (`subset.pair.*.tsv`).
|
153 |
|
154 |
+
Also, transcriptions are provided for each audio in:
|
155 |
- `data_offline_transcription.tsv` - transcriptions for **offline** examples and `keyword_path` from **online** scenario,
|
156 |
+
- `data_online_transcription.tsv` - transcriptions for the adversary, test examples from **online** scenario,
|
157 |
|
158 |
three columns are present within each file:
|
159 |
+
- `path_to_keyword`/`path_to_adversary_keyword` - path to the audio file,
|
160 |
- `keyword_transcription`/`adversary_keyword_transcription` - audio transcription,
|
161 |
- `keyword_phonetic_transcription`/`adversary_keyword_phonetic_transcription` - audio phonetic transcription.
|
162 |
|
|
|
166 |
- downloading the archive and constructing all the test cases based on the provided `tsv` files,
|
167 |
- `datasets` package.
|
168 |
|
169 |
+
In the latter case, the following should work:
|
170 |
+
```
|
171 |
+
load_dataset(path="voiceintelligenceresearch/MOCKS", name="en.LS-clean", split="offline")
|
172 |
+
```
|
173 |
|
174 |
The allowed values for `name` are:
|
175 |
- `en.LS-{clean,other}`,
|
|
|
220 |
|
221 |
The MOCKS testset was created from LibriSpeech and Mozilla Common Voice (MCV) datasets that are publicly available. To create it:
|
222 |
- a [MFA](https://mfa-models.readthedocs.io/en/latest/acoustic/index.html) with publicly available models was used to extract word-level alignments,
|
223 |
+
- an internally developed, rule-based grapheme-to-phoneme (G2P) algorithm was used to prepare phonetic transcriptions for each sample.
|
224 |
|
225 |
The data is stored in a 16-bit, single-channel WAV format. 16kHz sampling rate is used for LibriSpeech based testset
|
226 |
and 48kHz sampling rate for MCV based testset.
|
227 |
|
228 |
+
The offline testset contains an additional 0.1 seconds at the beginning and end of the extracted audio sample to mitigate the cut-speech effect.
|
229 |
+
The online version contains an additional 1 second or so at the beginning and end of the extracted audio sample.
|
230 |
|
231 |
The MOCKS testset is gender balanced.
|
232 |
|