Using Natural Language Processing to Support Interview Analysis
Dirk Riehle · 2013
Summary Highquality qualitative work is difficult and can be replicated by another researcher with access to the same data within the maximum level of interhuman agreement. Concept extraction involves identifying key topics within text and sentiment analysis may be used to identify the valence of terms associated with the concepts. Here, a nlp tool is used on interview data to explore if concept extraction and sentiment analysis can be used to validate or support coding by researchers. Work Results ● Literature review ○ Sentiment analysis ○ Agreement levels in data coding: human/human and human/machine ○ Qualitative data analysis, in particular interview analysis, especially formal/automated methods of demonstrating the validity of the results ● Preparation ○ Evaluate existing Open Source sentiment analysis tools and/or libraries ○ Identify sources of data for training domain knowledge and document the process ● Research results ○ A nlp tool capable of training on a corpus of public or private data and identifying key concepts from interviews and abstracting them ○ A theory on the validity of the approach of applying nlp to confirm human qualitative analysis