Regresar

Automating the measurement of heterogeneous chatbot designs

Abstract:

Chatbots are being increasingly used to provide a natural language interface to all kinds of software services. However, while there are many platforms and tools for chatbot development, they typically lack support to statically measure properties of the designed chatbots, as indicators of their size, complexity, quality or usability, and facilitating comparison. To attack this problem, in this paper we propose a suite of 20 metrics for chatbot designs. The metrics are defined on a neutral chatbot design language, becoming independent of the implementation platform. We have developed a tool, called Asymob, which supports the translation of chatbots defined in several platforms into this neutral format to perform the measurements. As a proof-of-concept, we evaluate the metrics over a collection of Dialogflow and Rasa chatbots from several sources and open-source repositories. Our metrics helped detecting quality issues statically, and served as a basis for comparing chatbots from different origins and built using different technologies.