Network text analysis of conceptual overlap in interviews, newspaper articles and keywords

We address the relative value of three information sources: costly interviews conducted in the field, newspaper articles that mention the areas in which the interviews took place, and keywords used to index the newspaper articles. Our research questions concern: (1) whether there is overlap in the information obtained from each source and (2) how the three information acquisition-extraction strategies employed can inform one another. This research project uses network text analysis as a framework for a mixed method approach to knowledge discovery. We show that concepts as well as the network structure of information obtained from interviews may be almost completely covered by networks representing the information extracted from a large number of news articles from a wide variety of sources, while the information overlap of interviews with article keywords was less straightforward. We also show how a conceptual network constructed from a small number of interviews can be used in a semantic pattern search that localizes interview topics in a larger network of news article topics. This approach thus uses newspaper articles to frame and elaborate the narratives of interviews in a larger cultural context.