Quality control

This is just a simple statement of how I am controlling the quality of transcriptions, etc. The information is partially redundant with the workflow for processing texts.

Orthographic Transcription

  1. Initial transcription by language helper
  2. Check by PI to see that all words & affixes are included. Places for revisions noted.
  3. Re-check by language helper.
  4. Second check by PI, checking for the same criteria.
  5. PI and language helper sit down together to resolve any outstanding differences.

The idea here is that the native speaker does what he does best, i.e., hearing the sounds, and the non-native speaker does what he does best, i.e., attending to every detail in the recording because he doesn't have a perceptual system or grammar optimized for parsing the important information out of the mess of words. Initially there was another round of checking and re-checking, but I found that my language helper made hardly any changes to the file at that point, so it wasn't worth it. (This is just to say that after the first round neither of our judgments changed, not that he wasn't doing it right. Sometimes he wasn't noticing the words I was hearing (especially function words), sometimes I was hearing it wrong.)

This all takes place in ElanCheck.

Phonetic Transcription

  1. PI and language helper sit down together to transcribe the texts word-for-word, based on the orthographic transcription.
  2. PI checks transcriptions against the recordings. Places for revisions noted.
  3. PI and language helper sit down together to revise.

I was personally surprised, having done a few texts, at how few revisions were required (step 3). This has proved a much less fraught process than the orthographic transcription did.

This all takes place in Language Explorer.

All contents copyright © 2017 Adam Baker, except where otherwise noted.