Blind Transcription

Due to the subjective nature of phonetic transcription (where measurement and verification of all relevant segments may not be feasible), Phon has a built-in system for performing multiple-blind transcriptions.

In a multiple-blind transcription process, records are transcribed by any number of separate users, who do not see each other's transcriptions. These separate transcriptions may then be resolved by an objective party, who can choose one over the other or compare them to reach a consensus transcription.

A typical example of multiple-blind transcription in Phon may proceed as follows:

  • Two users perform separate blind transcriptions of all records in one session (see Open session as transcriber (blind mode) for instructions).
  • A team of two different users review both sets of blind transcriptions together, while listening to each record.
  • For each record, this team does one of three things:
    • chooses the best of the two transcriptions
    • combines both transcriptions to reach a consensus
    • creates an alternate transcription which they agree upon (in extreme cases in which both blind transcriptions seem to be in error).

The transcription which is settled upon is entered into the IPA Target or IPA Actual field, and can be analyzed accordingly by project researchers, who will now have access to the best and most objective transcriptions possible. See IPA Validation for instructions on validating IPA Transcriptions. See also the inter-transcriber reliability report which provides several metrics of agreement between blind transcribers in a Phon session.