Air Reservation Conversational Agent

Abstract

Tourism is one of the largest industries in the world. The travelers generally plan their holidays in advance. They might need to select the airlines, book the seats, select the hotels, reserve the rooms and reserve travelling facilities from hotels to attraction. There are two main channels that are used commonly in order to fulfill the above requirements. They are online reservation web sites and travel agencies. In most of the cases travelers prefer travel agents over web sites because it is practically easier and they can get the tour arranged with minimal involvement. Further they tend to trust physically existing bodies than a virtual web site. So in order to provide customer service many travel agencies maintain customer care departments with quite large call center staff. Maintaining a large call centre requires a significant budget. On the other hand if the call center staff is reduced then customers have to wait until an agent becomes available. This is not a pleasant experience for customers and the agency might lose them. In order to find the appropriate tradeoff between operating cost and customer satisfaction, agencies have to think of alternate mechanisms that can serve the customers effectively while reducing budgets. Conversational Dialogue systems provide a solution for the above scenario. Currently there are many general purpose conversational agents but they are not capable of handling specialized domains such as the travel domain. This research effort focuses on building a generic conversational agent which can be used for flight reservations. Since there are many tools which can be used for converting text to speech and vice versa, the scope of this research has been limited to implementing dialogue parsing, dialogue management and dialogue generation based on English sentences. It is mainly driven using a natural language parser and a rule engine. Due to its low coupling architecture, the system can be extended to other domains as well.