Multiplatform Comparative Analysis of Intelligent Robots for Communication Efficiency in Smart Dialogs

Anna Pogoda, Ewa Lyko, Michal Kedziora, Ireneusz Jozwiak, Jolanta Pietraszko

Abstract: The objective of the paper is to analyze the algorithms and methods used in dialogue chatbot systems in terms of usability and quality of their functioning. The research used two knowledge corpuses differing in subject matter, number of specific intents, availability of entities, and number of training and test data. The research was performed on three platforms - RASA, Dialogflow, and IBM Watson. The influence of the number of intents and the presence of entities on the quality of the dialogue system using predefined metrics was examined. Additionally, the experiments concerned the analysis of chatbot operation for different ratios of training to test data. Moreover, the influence of the platform and corpus selection on the quality of operation and acceptance level of intent was examined.