You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
Make sure to clearly understand Type-A and Type-B requests, and the relavant limitations. Failing to follow the guidelines pertaining to the two acceptable types of RFWs will automatically lead to the disqualification of the RFW.
Take time to complete each section below with as much detail as is required to establish a comprehensive understanding of the underlying product specification.
Proposing a new platform for STT transcription work. Adding certain features that's lacking in Prodigy:
Organize users in Roles and Groups.
Reject a task and sent it back to the transcriber.
Generate report and track a task.
State Machine to handle task state.
Edit transcription before submit.
Is This Really Necessary?
We are facing problems with scaling up Prodigy. We are hoping that a custom software will curb some of the issues we have faced and make transcriber's job easier.
Motivation
In finding workarounds with Prodigy we have ended up creating up to 87 instances each for a group. Running a CRON job to query the database directly and upload to multile Google Sheets for reporting. We have also faced problem of duplicate task appearing thus wasting trasncriber and reviewer's time.
Named Concepts
We can use State Machine to represnt the state a task is in.
Possible states for a task are:
Imported : the starting state of a task.
Transcribing : a task has been assigned to a transcriber and is currently being worked on.
Trashed : an audio segemnt is not usable and is trashed.
Submitted : a work has been submitted for reviewed.
Accepted : A reviewer has reviewed and has accepted the task.
Finalized : The Second reviewer has finalized the transcription and has accepted the tak in finalized state.
Group : Every transcriber belongs to a group (eg. stt_tt_ga, stt_cs_gb ..)
Role : Role can be Transcriber or Reviewer, Final-Reviewer.
A user is assined a role.
Examples
No duplicate task allocation to mulitple annotators.
Report generation for users on a common platfrom.
Trasncriber can go back to a task and eddit multiple times before submitting.
Notify when a task has been rejected.
Conceptual Design
Drawbacks
Will require time and effort to build it from scratch.
Alternatives
Extend an existing Open source project like Lable-studio and fork it. Customize it for our needs.
New Data
We get a better idea of where a task is stuck in the pipeline. More analytics for transcriber performance.
Adaption Window
Four weeks.
The text was updated successfully, but these errors were encountered:
Table of Contents
Housekeeping
Make sure to clearly understand Type-A and Type-B requests, and the relavant limitations. Failing to follow the guidelines pertaining to the two acceptable types of RFWs will automatically lead to the disqualification of the RFW.
Take time to complete each section below with as much detail as is required to establish a comprehensive understanding of the underlying product specification.
ALL BELOW FIELDS ARE REQUIRED
Owner
@TenzinGayche @spsither
Summary
Proposing a new platform for STT transcription work. Adding certain features that's lacking in Prodigy:
Is This Really Necessary?
We are facing problems with scaling up Prodigy. We are hoping that a custom software will curb some of the issues we have faced and make transcriber's job easier.
Motivation
In finding workarounds with Prodigy we have ended up creating up to 87 instances each for a group. Running a CRON job to query the database directly and upload to multile Google Sheets for reporting. We have also faced problem of duplicate task appearing thus wasting trasncriber and reviewer's time.
Named Concepts
Possible states for a task are:
A user is assined a role.
Examples
Conceptual Design
Drawbacks
Will require time and effort to build it from scratch.
Alternatives
Extend an existing Open source project like Lable-studio and fork it. Customize it for our needs.
New Data
We get a better idea of where a task is stuck in the pipeline. More analytics for transcriber performance.
Adaption Window
Four weeks.
The text was updated successfully, but these errors were encountered: