The use of artificial intelligence and chatbots for recommendations of products to customers

The objective of this thesis was to create a prototype of a chatbot that imitates the conversation of a bartender, recommending whisky. Whisky was taken as an example, as a whisky recommendation bases on the individual preference and taste of the consumer. For similar products, such as wine, perfume, coffee etc., the chatbot would work as well. The literature identifies different filtering methods for recommender systems, such as collaborative filtering (e.g. Customers that have bought product A also liked product B), content-based filtering (e.g. You bought a printer, you might also need ink) or knowledge-based filtering (e.g. We know, you like skirts: Do you prefer the red or the blue skirt?). Two or more methods can be united into so-called hybrid systems, which try to combine the best aspects of each system. Chatbots are an up-and-coming technology, which help to create conversations. To find out about exact preferences of whisky drinkers, conversational recommenders can be used. Existing recommender systems combined with chatbots cannot act personally – for example they are not able to determine whether a guest is experienced or not in whisky drinking or if it is a returning or new customer. Furthermore, they cannot react individually on the feedback of the consumer. The thesis statement of the current work addressed this gap in the literature: It is possible to recommend products based on individual taste and preferences to customers by hybrid recommender systems (knowledge-based and content-based) combined with chatbots, which can distinguish between returning and new customers....