Arabic Dialogue System for Hotel Reservation based on Natural Language Processing Techniques

Abstract

In this paper, we present an Arabic dialogue system (also referred to as a conversational agent) intended to interact with hotel customers and generate responses about reserving a hotel room and other services. The system uses text-based natural language dialogue to navigate customers to the desired answers. We describe the two main modules used in our system: the parser and the dialogue manager. The parser is based on the Government and Binding theory. Customers can inquire about room availability, hotel services and negotiate a desired reservation. We report an experiment with 500 volunteers unfamiliar with the system in a real environment. The users were asked to interact with the system and then to judge the dialogues as “very bad,” “bad,” “neutral,” “good,” or “very good.” We found that 66.92% of the dialogues were judged to be “very good” and 92.3% were judged to be “good” or “very good”. These results confirm the viability of using an Arabic dialogue system to tackle the problem of interactive Arabic dialogues. Finally, we discuss future directions for enhancing our dialogue system with more sophisticated and intuitive interaction.