Pervasive human-AI collaboration in interactive tourism

Abstract

Tourism is one of the most important components of the economy for many countries in the world. Every year, hundreds of millions of tourists from all over the world choose certain countries for travel and book hotels in advance taking into account past experience. Almost every traveler whether or not travelling abroad faces a certain lack of information during the trip especially regarding historical and architectural monuments. And almost every tourist would like to have a personal guide who could tell about any object of interest to him in a new city. This personal guide is very expensive and not every tourist can afford its daily accompaniment. In this paper, we present a new way to identify cultural and historical objects and get their detailed description using a telegram-bot. Using a telegram bot, we have integrated an improved image retrieval system and ChatGPT to create a new way of interaction for the tourist. To get more detailed information on any tourist object, a tourist needs to take a photo of this object directly in the telegram bot. The telegram bot will give the original photo of the object, a URL link to the site of the object and immediately give a detailed description of the object generated by ChatGPT. After that, the tourist will be able to ask ChatGPT any questions immediately in the telegram bot, including details of the story or something else. In this work we were using image augmentation for image retrieval to improve the accuracy. Deep convolution neural network was used as a backbone. As a result of the experiments conducted in the city of Tashkent, it was proved that the proposed way of interaction for tourists is feasible and can be used in many other applications.

Date of Award	2023
Original language	English
Awarding Institution	The Hong Kong University of Science and Technology
Supervisor	Pan HUI (Supervisor) & Tristan Camille BRAUD (Supervisor)