Enhancing Disaster Resilience Studies: Leveraging Linked Data and Natural Language Processing for Consistent Open-Ended Interviews

Abstract

Researchers have long focused on disaster resilience to mitigate calamity disruption. Disaster resilience is a complex and multi-faceted concept that is challenging to measure. Quantitative methods have traditionally been used to assess disaster resilience, but a growing interest in qualitative methods like open-ended interviews has emerged to understand experiences and perspectives. To gain deep and consistent knowledge, an open-ended interview should focus on an interviewee’s point of view and ask follow-up questions from a knowledge base that consists of relevant information; otherwise, this can lead an open-ended interview to deviate from the interviewee’s point of view to the interviewer’s point of view. In contrast to what is desired, individual interviews with last year's students in the field of civil engineering with a predefined and limited knowledge base demonstrated inconsistency in asking a follow-up question from an already existing open-ended interview. To tackle this gap, firstly, we suggest a knowledge base that can be built from peer-reviewed papers published in the disaster resilience field; secondly, we suggest a Natural Language Processing based Decision Support System using Sentence Embedding that can analyze the interviewee’s response and find resources from the knowledge base to assist the interviewer in making a consistent follow-up question