Automatically generating decision-support chatbots based on DMN models
Bedilia Estrada-Torres, Adela del-Río-Ortega, Manuel Resinas
TL;DR
This work tackles the challenge of building decision-support chatbots by automating the transformation of DMN decision models into fully functional chatbots. The authors introduce Demabot, a low-code toolchain that automatically extracts DMN elements, maps them to chatbot components (entities, intents, training phrases, contexts, actions), generates training phrases with an NLG-based approach, and integrates with Dialogflow to deliver a user-facing interface. Key contributions include a principled DMN-to-chatbot mapping, an optimization-driven conversation manager to minimize user interactions, and an empirical evaluation with 15 participants demonstrating correctness, usability and reduced development effort. The approach enables domain experts to create domain-specific chatbots with limited technical knowledge, supporting broader adoption of decision-support chatbots across industries while highlighting areas for future NLP enhancements.
Abstract
How decisions are being made is of utmost importance within organizations. The explicit representation of business logic facilitates identifying and adopting the criteria needed to make a particular decision and drives initiatives to automate repetitive decisions. The last decade has seen a surge in both the adoption of decision modeling standards such as DMN and the use of software tools such as chatbots, which seek to automate parts of the process by interacting with users to guide them in executing tasks or providing information. However, building a chatbot is not a trivial task, as it requires extensive knowledge of the business domain as well as technical knowledge for implementing the tool. In this paper, we build on these two requirements to propose an approach for the automatic generation of fully functional, ready-to-use decisions-support chatbots based on a DNM decision model. With the aim of reducing chatbots development time and to allowing non-technical users the possibility of developing chatbots specific to their domain, all necessary phases for the generation of the chatbot were implemented in the Demabot tool. The evaluation was conducted with potential developers and end users. The results showed that Demabot generates chatbots that are correct and allow for acceptably smooth communication with the user. Furthermore, Demabots's help and customization options are considered useful and correct, while the tool can also help to reduce development time and potential errors.