Skip to content

Latest commit

 

History

History
49 lines (38 loc) · 2.27 KB

QA Tool Guideline.md

File metadata and controls

49 lines (38 loc) · 2.27 KB
Raw

aiXplain QA Tool Guideline

Site: : Dataset name: English Username: <your username> Password: <your password>

Instructions:

  1. Accessing the Tool:

    • Open the provided URL in your web browser.
    • Navigate to the "Intro Annotator" page.
    • Login using the provided username and password.

  2. Dataset Selection:

    • After logging in, go to the QA page.
    • On the sidebar, select the dataset that you have been assigned to.

  3. Annotation Process:

    • The samples for annotation will be presented to you in a specific order.
    • Start with the sample having the highest Word Error Rate (WER).
    • For each sample, follow the steps below:

    a. Listening to the Audio:

    • Click on the play button to listen to the audio recording.

    b. Selecting the Transcription:

    • Choose the transcription that best matches the audio.
    • Ensure that the transcription accurately represents the content of the audio.
    • If necessary, you can make post-editing changes to the selected transcription.

    c. Discarding Poor Quality Audio:

    • If the audio is not of good quality, it should be discarded.
    • Select one or more reasons for discarding from the following options:
      • Repetition: The audio contains repeated sentences.
      • Incorrect Accent: The accent in the audio is not appropriate.
      • Incorrect Pronunciation: The audio has incorrect pronunciations or includes "hmm" sounds.
      • Poor Recording Quality: The audio has background noises, lips sounds, or is of low quality.
      • Incorrect Pauses: The pauses in the audio are not appropriately placed.
      • Incorrect Speed: The audio is either too fast or too slow.
      • Inconsistent Voice: The voice level in the audio is inconsistent and does not feel natural.

    d. Providing Feedback (Optional):

    • If you discard an audio sample, you can provide additional feedback regarding the reason for discarding.

    e. Submitting the Annotation:

    • Once you have selected the best transcription or discarded the audio, click on the "Submit" button to save your annotation.

Note: Please ensure that you follow these guidelines while using the QA Tool for checking recordings used in TTS model training.