Thanks to visit codestin.com
Credit goes to www.scribd.com

Scribd
Upload
12 views3 pages

Step 4 Data Factory Platform Guide

The document outlines the procedures for transcribers (TXs) and quality assurance (QAs) personnel on a transcription platform, emphasizing the importance of claiming tasks as assigned, adhering to transcription rules, and ensuring accuracy in timestamps. It also highlights the need to filter out invalid audios, correct minor mistakes, and follow specific guidelines for spell checking and reworking tasks. Overall, the document stresses that accurate transcription is crucial for the success of the project and the AI model training.

Uploaded by

Salah eddine Khayat
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
12 views3 pages

Step 4 Data Factory Platform Guide

The document outlines the procedures for transcribers (TXs) and quality assurance (QAs) personnel on a transcription platform, emphasizing the importance of claiming tasks as assigned, adhering to transcription rules, and ensuring accuracy in timestamps. It also highlights the need to filter out invalid audios, correct minor mistakes, and follow specific guidelines for spell checking and reworking tasks. Overall, the document stresses that accurate transcription is crucial for the success of the project and the AI model training.

Uploaded by

Salah eddine Khayat
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 3

Please read thoroughly until the very end!

Process:

Platform link http://nls-df-inner.aliyun-inc.com/

Only claim tasks assigned in the group notice. Please do not claim order
ids that we did not mention, otherwise no settlement will be made!!!

For TXs (Transcribers)

1. Ensure that you correctly apply transcription rules in the transcription process.

2. Claim 1 task ->finish it and submit -


>claim 1 new task. Do not reserve tasks. You will not run out of tasks to do. You
will only delay the process if you claim many tasks all at once.

3. Transcribe the valid/clear parts and transcribe according to the rule/guidelines.


English sentence is also valid in Malay transcription project, pls transcribe it.

4. Always choose the correct labels.

5. TIME STAMPING, if you can transcribe an entire audio, adjust the timestamp ac
cordingly/ properly. If there are words at the beginning or ending of the transcript
ion that you cannot transribe, adjust the timestamp to cut them.

For QAs (Quality Assurance)

1. Claim 1 task ->finish it and submit -


>claim 1 new task. Do not reserve tasks. You will not run out of tasks to do. You
will only delay the process if you claim many tasks all at once.

2. No need to review invalid audios, remember to filter out valid audios first in each
task. see screenshot below for detailed steps.
3. You need to edit out the mistakes and make the necessarry corrections if only mi
nor mistakes. Click the blue button (correct, next) and click "pass, submit" under
the last audio, and continue to the next task. (if you find that serious mistakes an
d the overall quality of this task if poor, leave your comment in the "message boar
d" and click the red button for one audio, finally return the task to annotator)

4. Make sure to adjust the timestamp accordingly.

If you change one audio from valid to invalid, refresh the page to update the au
dios since you filtered the VALID audios only.

Spell checker on platform

The spell checker is for COMMON American English words. The system may or may not c
onsider a word mispelled specially for Brand names, Person's name, Company name, and
other uncommon words.

Refer to HFW in the group link (High Frequency Words) or Google for proper word spellin
g/ format.

Before you submit the task, make sure you go through all audios with "请请请
" to double check the misspelled word/s.

The OOV in RELATED INFORMATION box will show the mispelled word/s, NOT t
he correct spelling of the misspelled word/s.
OOV will not update even if you already corrected the spelling.

Reworking

 Rework orders are orders that didn't pass the client's acceptance

 Tasks will be returned to your checking tab if you are required to do the reworkin
g tasks

 Wait for the rework report (if there's any) or further instructions before reworking
.

All those results will be used to train the AI model, if your transcription is inaccurate, the
AI will be the same. Your transcription will be useless to the client, and it WILL affect th
e entire project.

Valid duration will only be concluded when the order/task is PASSED by the Client's acc
eptance.

You might also like