Construção de chatbots AIML com a ajuda de uma ferramenta de modelagem visual baseada na linguagem BPMN

Abstract

A chatbot is a computer program capable of maintaining a dialogue with a human being. Among the various existing techniques for building chatbots is an open specification based on Extensible Markup Language (XML) called Artificial Intelligence Markup Language (AIML). Designing dialogue flows for chatbots is a process that depends on the technology chosen for its construction. Building these flows in AIML requires knowledge in tags. Dialogue flows is a difficult and error-prone task. Visual modeling for the construction of flows by non-programmers could bring agility to this process. There is a visual notation for modeling processes called Business Process Model and Notation (BPMN) that has good acceptance both among professionals who are more distant from technology and for professionals in the field. It was proposed in this dissertation a visual approach to define flows with the help of BPMN and an algorithm to convert it into AIML. The proposal is to specify the dialogue flow of a chatbot, with the help of BPMN, and to transform it into AIML automatically, allowing users without knowledge in AIML and to be able to design them intuitively. To confirm this assumption, a converter, called BPMN2AIML, was built, in addition to a chatbot, called ARI, which incorporates it. Some dialog flows were mapped, then the diagrams were converted to AIML and loaded into the chatbot. The converter was evaluated by 12 botmasters who built several chatbots according to the proposed business problems, and found the experience satisfactory and useful.