Enhancing Conversational Troubleshooting with Multi-modality: Design and Implementation

Conversational troubleshooting is an increasingly popular technology that consists in utilising dialogue agents to support users of a system through a conversation-based question-answering process, typically through a chatbot. Despite their widespread use, current troubleshooting technologies lack a thorough integration with the applications on which they are overlaid, as they rely only on text to help the users. In this paper, we present TINI, an open-source conversational troubleshooting tool that is multi-modal and multilingual, relies on simple configuration files, and is ready to be deployed in web applications. Users can ask questions to the conversational agent explaining the issue faced; the system analyses it together with the interaction’s context to locate the root problem. Finally, it proposes a solution which engages the user multi-modally: with text in the chat and hints in the graphical interface. A table-based configuration improves system maintainability and enables dialogue designers and field experts to work on the conversation without any coding experience required.