There are many methods for ensuring transcription accuracy. For example, a concordance assessment involves two typists independently transcribing an interview and then comparing the two transcripts. However, it is rare that a researcher has the resources for acquiring the time of two typists for one transcript.
Ideally a researcher would have access to as much time and personnel as is necessary to ensure data accuracy and integrity. However, a researcher must balance the ideal with practical considerations of time and funding limits.
Thus, an optimal strategy is typically used wherein each interview is transcribed by a professional and then proofread by the interviewer. Research projects involving intensive data collection, however, often delegate this review process to another team member other than the primary interviewer.
Training Transcription Proofreaders
The training of proofreaders should be modified to fit the research project and will vary with research structure, setting, type and volume of data, and the analytical approach taken. Discrepancies can arise if a proofreader is not completely familiar with transcription protocols, the research project and related terminology or vernacular used by the interviewee.
These individuals must, at the very least, be familiar with protocol for processing “raw” and “prepared” data, apply this protocol consistently and know what to do when exceptions arise.
Above all, transcripts must be readable and accessible to collaborators. This means avoiding linguistic- or phonetic- type transcripts or producing a simplified set of transcription symbols to indicate the precise length of pauses, verbal stresses, overlapping talk, and interviewer or analyst descriptions.
Inevitable Transcription Errors
All transcribers will leave out words here and there or recount phrases not exactly as spoken. Thus, establishing a review protocol and reviewing all transcripts is a significant dimension of data preparation. This protocol will include instructions for transcribers and proofreaders to indicate when sections of audio recordings are inaudible by including the phrase “inaudible text segment” and indicate overlapping speech that cannot be deciphered with the phrase “cross talk.”
Reviewing by Subset
If only a subset of transcripts in a database is to be reviewed then it is imperative that these transcripts are reviewed with heightened scrutiny. Failure to do so may leave transcription errors unnoticed until analysis is underway, at which point it is time-consuming to correct errors. Such scrutiny includes checking transcripts against the original audiotape.
Time markers, timestamps are utilized when this protocol is activated. At the time of each transcript page a time-stamp indicates the point in the recording that the page transcribes. This allows for quick retrieval of specific audio components for comparison with textual data.
Source: McLellan, Eleanor, MacQueen, Kathleen M., and Neidig, Judith L. 2003. “Beyond the Qualitative Interview: Data Preparation and Transcription.” Field Methods 15(1): 63-84
My roommate was talking about looking into production transcripts. She was trying to learn a little more about how they are made and perfected. It is nice to know that they can be reviewed by a subset.