2012 MRes projects
Find a SECReT supervisor
prism apply now

Automating the conceptual analysis of large-scale text-based subjective data sets

21 March 2013

Ingolf Becker

A common starting point of research in social sciences is grounded theory analysis. Through the use of some initial qualitative data-gathering, such as questionnaires and interviews, concepts and theories are discovered. As this analysis is done by hand, there is a trade-off between the time-consumption of the analysis and the completeness of concepts found. Using computational tools for this analysis can enable the researcher to gather a more comprehensive picture with little effort.  The aim of this dissertation is to develop and test a framework that allows for automatic conceptual analysis of subjective subject-specific data. The initial research hypothesis follow this aim: Does the automatic computational analysis reveal the same/more insight into the security issues at the organisation than the manual approach? Is it possible to infer some underlying sentiment to each of the interviews? Can this be correlated to other features of the interview? The research group around one of my supervisors, Angela Sasse, has conducted interviews with employees of two organisations on the topic of security. The corpus consists of 118 transcribed 30-40 minute interviews from company A and a slightly smaller set of interviews from company B, about 400,000 words in total. The data has been anonymised. We hope that the outcome of this research will allow a more accurate and foremost less time consuming approach to the analysis of this initial research. As interviews are used throughout security science to gather data, it will be possible to transfer the methodology developed here to other areas.