laboroai / laborotvspeech Goto Github PK

View Code? Open in Web Editor NEW

83.0 83.0 11.0 494 KB

License: Apache License 2.0

Shell 100.00%

laborotvspeech's People

Contributors

Stargazers

Watchers

Forkers

eiichiroi ishine double22a sciai-ai houwenxin narakai nyxrobotics tbright17 lan2016286 nayanjha16 splinter21

laborotvspeech's Issues

Too long utterances in the dev set

I found that the dev set includes too long sentences.
I suggest removing such long sentences for the official development data.
My suggestion is to remove utterances for more than 30 seconds (e.g., remove the top 33 utterances from the following list).

$ sort -k 2 -nr dump/raw/dev/utt2num_samples | head -n 34
v001_dev_RlFGUdfe 2659360
v001_dev_tVD60s3X 2603200
v001_dev_XcWLZ8Ds 2284800
v001_dev_m3w23nXF 1494400
v001_dev_yhU29kG8 1464960
v001_dev_yIo4ACN1 1390720
v001_dev_rUVXkS0Z 1346560
v001_dev_ebeV3qs8 1223840
v001_dev_PPdlAyGz 1205120
v001_dev_6JZPzq4k 1130560
v001_dev_g8vWy3VB 987040
v001_dev_BYca8AqR 982560
v001_dev_kpypOOpB 966240
v001_dev_fwWvJhdZ 939840
v001_dev_YVVAj1Ml 939520
v001_dev_m9vLDdIr 743520
v001_dev_tvigUeHB 694080
v001_dev_puL8hGoC 689120
v001_dev_eqpJcjYj 681920
v001_dev_IroOq3x7 673920
v001_dev_tbyYEkw2 562240
v001_dev_gt3MQcrd 560160
v001_dev_Xkf0hGjo 527200
v001_dev_yIkO2OlI 522560
v001_dev_xL5OF2nG 515680
v001_dev_dfcYXj1r 506880
v001_dev_YUtqHgLy 504800
v001_dev_bJ4u7YZd 500800
v001_dev_I7fBUbhG 497600
v001_dev_6lwD3FOm 487680
v001_dev_hEaNS9HA 486240
v001_dev_uGKSJThT 483200
v001_dev_5SMfe193 480160
v001_dev_g51KQbqF 479200

What speech-to-text software can I use to easily try oscar_3gram_lm_v1.0.tar.gz?

I want to implement offline Japanese speech to text.
I have set up Kaldi, but I don't know how to run Kaldi by specifying "oscar_3gram_lm_v1.0" which is distributed here.
If you don't mind, could you give me a sample shell script or something?

Error during `local/tedx-jp/10k_data_prep.sh`

When I perform local/tedx-jp/10k_data_prep.sh, it was failed with the following error

local/tedx-jp/10k_data_prep.sh: All required files exist in tedx-jp.
local/tedx-jp/10k_data_prep.sh: Creating a directory containing all valid subtitle segments.
2020-11-19 20:25:05,588 (make_raw_dir:235) INFO: done making data/tedx-jp-all_raw
fix_data_dir.sh: kept all 273 utterances.
fix_data_dir.sh: old files are kept in data/tedx-jp-all_raw_whole/.backup
patching file data/local/tedx-jp-10k_verbatim.tmp/text_content
Hunk #90 FAILED at 2657.
1 out of 320 hunks FAILED -- saving rejects to file data/local/tedx-jp-10k_verbatim.tmp/text_content.rej

It seems that the patch file would be wrong.
Could you let me know how we solve it?

English and Music in TEDxJP 10k corpus

The TEDxJP 10k corpus contains some inapporopriate data for evaluation of Japanese speech recognition.

In following videos, Japanese people talk in English.
But the corpus uses subtitles that are automatically translated into Japanese.

Then, following videos contain some music (corpus uses the interval of music).

It may be better to be removed from the corpus.

laboroai / laborotvspeech Goto Github PK

laborotvspeech's People

Contributors

Stargazers

Watchers

Forkers

laborotvspeech's Issues

Too long utterances in the dev set

What speech-to-text software can I use to easily try oscar_3gram_lm_v1.0.tar.gz?

Error during `local/tedx-jp/10k_data_prep.sh`

English and Music in TEDxJP 10k corpus

Recommend Projects

React

Vue.js

Typescript

TensorFlow

Django

Laravel

D3

Recommend Topics

javascript

web

server

Machine learning

Visualization

Game

Recommend Org

Facebook

Microsoft

Google

Alibaba

D3

Tencent