Exploratory text data analysis for quality hypothesis generation

Freestyle text data such as surveys, complaint transcripts, customer ratings, or maintenance squawks can provide critical information for quality engineering. Exploratory text data analysis (ETDA) is proposed here as a special case of exploratory data analysis (EDA) for quality improvement problems with freestyle text data. The EDTA method seeks to extract useful information from the text data to identify hypotheses for additional exploration relating to key inputs or outputs. The proposed four steps of ETDA are: (1) preprocessing of text data, (2) text data analysis and display, (3) salient feature identification, and (4) salient feature interpretation. Five examples illustrate the methods.