Abstract

Audio data is a valuable and underutilized source of information at the IAB. Surveys that use interviewers (e.g., the IAB PASS survey, the IAB Quality of Life and Participation survey, and the IAB Establishment Panel) routinely record significant portions of the interviews for quality control purposes. However, analyzing these recordings is labor-intensive and costly, so analyses are conducted only selectively and to a minimal extent. Automating the analysis of such audio recordings offers enormous potential for improving data quality, increasing the efficiency of survey processes, and generating new metadata and analysis data. As part of a pilot study, we therefore plan to transcribe the recorded interviews from PASS Wave 18 and evaluate the obtained text data with a focus on data quality (measurement error).