A Model-Driven Approach Employing DSL and Generation Templates to Accelerate the Development of Conversational Agents for Smart Tourism

Conversational agents are intelligent, ubiquitous software applications widely used across various sectors, such as commerce, tourism, and more. Their key benefits include automating tasks, improving customer service, and ensuring service availability. The development of conversational agents utilizing artificial intelligence (AI) techniques represents a significant advancement in Natural Language Processing (NLP). Numerous studies employ deep learning and NLP methodologies to construct sophisticated conversational agent systems. Additionally, developers and companies often utilize APIs provided by intent recognition services like Dialogflow and Amazon Lex to easily create conversational agents using graphical forms, which enhance conversational agent functionality. However, these APIs have limitations, such as potential dependency on specific NLP service providers and associated high costs. Besides, the lack of a specialized conversational agent development platform for the tourism domain poses a considerable challenge. To address these limitations, this work tackles critical gaps in conversational agent development tools by constructing a graphical Domain-Specific Language (DSL) and code generation templates for accelerating the development of conversational agents tailored to smart tourism’s needs. First, we provide a designed metamodel to define the abstract syntax of a DSL. Second, we implement the metamodel using the Eclipse Modeling Framework. Third, we develop a graphical interface that incorporates intuitive icons to simplify the creation of conversational agent models. Fourth, we define code generation templates to translate the graphical models into executable agent source code. Finally, we validate the proposed approach to demonstrate its effectiveness and applicability in real-world scenarios, reducing development time and avoiding the costs associated with NLP services.