Extracting high-quality end-user requirements via a chatbot elicitation assistant

Abstract

Requirements elicitation can be done using several methods, depending on what segment of users one would want to target. This thesis studies whether a chatbot can be used to elicit high-quality end-user requirements. This thesis introduces ReqBot, a prototype chatbot elicitation assistant; we investigate ReqBot's impact on the quality of requirements and its usability. Quality is represented by referential ambiguity and vagueness, a subset of ambiguity, which in turn is a quality metric for requirements. ReqBot aims to detect these ambiguities and resolve them through interaction with the user. ReqBot is tested against a Google Form, a representation of a current method of end-user requirement elicitation. While both gained a high usability score, possibly due to the simplicity of the user input interface, there was no significant difference in usability between the two. ReqBot was able to detect and resolve ambiguities with a decent recall, however, it still lacks in precision.