Process

Samples used for reliability

The samples should be copied here and the transcription for reliabiliy and rely output can be found here

What needs to be edited?

We decided it should be at least 10% per each timepoint for VISTA samples for both trained and untrained topics, separately in each language (but see more information re: decision, below)
There are six trained and two untrained topics per bilingual participant and currently we are transcribing only pre 1, pre 2, post 1, post 2, 3mofu (for older cases, only), 6mofu, 12mofu
- You can find the training condition here https://app.smartsheet.com/sheets/29c56pPH98JW66r6h7fxp8M3rM5cv6xQJqPJm4m1
- This should result in one trained and untrained topic being double transcribed at each timepoint within a participant (ends up being 16.7% for trained topics per participant in each language at a timepoint and 50% for untrained topics).
  - It doesn’t matter if the topic is the same or different across languages that are being transcribed
  - We have to ensure we assign the reliability transcription to a transcriber who did NOT complete the original transcription

Copy and blind the audios following the guidance above and ensure there is a key so we can reference it @Andrew Collins
If we have a Whisper transcription for the sample, blind the sample and copy it into the reliability folder @Andrew Collins
Copy and paste from the MB datasheets, the final scripts so that the reliability transcribers (and future first transcribers), know what the target is. They should just reference this prior to transcribing so they are aware of the topic and general content @Andrew Collinsyou can direct @Rylee Manning to assist with this.
@Sonia Marqués is blinding the Smartsheet and creating a report so that the data can be entered to the VISTA Smartsheet

We use this reports to assign samples for reliability and also keep the progress: