Handling Uncertainty in the Development of Conversational Assistants with RELAXed User Stories

Conversational assistants have become increasingly popular, being used in different application domains, from Smart Homes to Health Care. Nevertheless, a considerable challenge concerning their development is to deal with their intrinsic uncertainty. Here we want to focus on the development early stages of this kind of system. Thus, to address uncertainty and simultaneously keep the documentation easy to maintain and user-centered, we opted to specify the requirements by combining the RELAX approach and user stories, creating the RELAXed User Stories approach. To accomplish this, we started by specifying a metamodel that expresses the concepts and relationships between user stories and RELAX. Also, a template to specify a RELAXed User Story is proposed. The main contribution is to help the requirements engineer work to specify conversational assistants, where uncertainty plays a significant role, in particular in the agile development context where user stories are common artifacts. We illustrate the use of RELAXed User Stories with a simple example. Finally, we performed a preliminary qualitative evaluation of the approach, with encouraging results.