University of Twente Student Theses


Knowledge Graph Driven Conversational Virtual Museum Guide

Liu, Dou (2021) Knowledge Graph Driven Conversational Virtual Museum Guide.

This is the latest version of this item.

[img] PDF
Abstract:Conversational agent in museum is studied and developed during the past 20 years, but it is still underexplored for museum conversational agents to utilize knowledge graph to guide the visitors and share knowledge with them. In this thesis, we addressed how to create a knowledge graph driven conversational virtual museum guide, as well as how to utilize the knowledge graph to develop its skills, such as Q&A and recommendation, aiming to share information with the visitors when they are visiting a virtual exhibition and help them explore the knowledge graph through dialogue. To address the aforementioned questions, the solution is composed of following aspects. A culture heritage knowledge graph, containing three thousand paintings and other entities related to the paintings, is designed and built by extracting relevant entities and relationships from Wikidata. Then, we constructed our conversational virtual museum guide with Dialogflow, where six functions are designed to support different kinds of tasks and conversations. Skills of the agent, such as Q&A and recommendation, are implemented as webhook service, i.e. backend service. The Q&A utilized a template-based method to identify the question types, generate the query of the question and fill the answer fetched in the slot of response template. The recommendation give suggested entities in the knowledge graph by combining the user's interest and similarity scores calculated using node embedding. Finally, our knowledge graph driven museum conversational guide is integrated with a virtual museum exhibition and can be visited by the public. As for the evaluation, an online user study was carried out with 10 participants, which included two stages. The first was playing through the virtual museum and finishing given tasks with the conversational virtual museum guide, and the second stage was completing a questionnaire about the experience with the conversational guide. The questions related to the usability and functionality of the conversational agent are designed to measure its performance, which helps to validate if the conversational agent can have natural conversation and share knowledge as a museum guide. The result shows that most users agree that having conversations improves their experience, and they are satisfied with the functions of the agent, especially the recommendation. Our research shows the possibility to integrate knowledge graph with museum conversational agent. Features driven by knowledge graph can improve the experience of visitors.
Item Type:Essay (Master)
Faculty:EEMCS: Electrical Engineering, Mathematics and Computer Science
Subject:54 computer science
Programme:Computer Science MSc (60300)
Link to this item:
Export this item as:BibTeX
HTML Citation
Reference Manager


Repository Staff Only: item control page