Automated coding of qualitative interviews with latent semantic analysis
Coding and analysing qualitative interviews is one of several core techniques used in marketing research. Qualitative methods offer valuable information hardly gained by standard quantitative methods since open-ended questions and interviews provide deeper insight into customer demands. The main disadvantages of qualitative methods are their inherent subjectivity and their high costs. We tackle this problem by applying latent semantic analysis (LSA) in a fully automated way on transcripts of interviews and we propose two algorithms based on LSA. We evaluate the algorithms against two separate real-life cases taken from the automobile industry and from the Austrian mobile phone market. Thereby, we compare the machine results against marketing expert judgements and show that the algorithms proposed provide perfect reliability with appropriate validity in automated coding and textual analysis.
Full Text: PDF