Over the last years, user feedback has become a valuable source for requirements elicitation. Nevertheless, limited feedback quality and lack of contextual information force product teams to engage in time-intensive conversations with users to understand the feedback and make it actionable. Chatbots can speed up such conversations by detecting ambiguities in user statements and classifying missing information needed. This paper introduces a chatbot that supports product teams in eliciting contextual information lacking in feedback. We implemented the chatbot using the open-source framework Rasa. Through a case study, we tailored and trained the chatbot with publicly available app store reviews from a large video streaming app, improving the chatbot in three iterations. A preliminary evaluation with 23 test users showed that our design could elicit three times more contextual information than traditional form-based feedback. By providing a structured feedback report at the end of the conversation, we significantly decreased the time and resources needed to review user feedback. Even though the chatbot needs additional training iterations and evaluations in operational settings, the initial results are promising.

Keywords: Chatbot, Contextual Information, Requirement Elicitation, User Feedback, Rasa Open Source

A demonstration of the chatbot can be watched here: https://youtu.be/a2gSBaijiY8