Please use this identifier to cite or link to this item: http://dspace.lib.uom.gr/handle/2159/28676
Title: Question answering system based on tourism knowledge graph
Other Titles: Σύστημα Ερωτοαπαντήσεων βασισμένο σε τουριστικό γράφο γνώσης.
Authors: Μανώλη, Χριστίνα
Keywords: Knowledge Graphs
Question Answering System
Natural Language Processing
BERT Embeddings
Issue Date: 2023
Publisher: Πανεπιστήμιο Μακεδονίας
Abstract: Nowadays, our society finds itself in the midst of a great evolving period including the important research that is being conducted towards important and useful technologies. Amongst those technologies we can find Knowledge Graphs (KG) and Question Answering Systems (QASs). In this paper, we conduct a thorough and extensive research on both Knowledge Graphs and Question Answering Systems in order to highlight their potential and ultimately develop our own QA System that employs our original, purpose-built (domain specific) Knowledge Graph for Tourism. This study mainly consists of two parts. Firstly, we analyze the theoretical background in order to better understand our technologies and set the fundamental terms and tools that we are going to later use on the second more technical part of the thesis. First and foremost, we introduce the concept of Knowledge Graphs; graphs of interconnected data which consist of real-world entities of interest, the relations that interconnect them and the attributes that help better describe them. From there we move on to duly note the KG background information; from their historical Background to their significance and various applications. Moreover, we extend this literature review into Question Answering Systems as well. Additionally, we break down any related research that has been conducted around the field of Tourism Knowledge Graphs. Regarding the second part of this thesis, we start by analyzing the process of creating our Tourism Knowledge Graph for the case study of Santorini. Furthermore, we present in detail how we built our Question Answering System by deploying a Template-Based model and implementing technologies like Natural Language Processing and state-of-the-art BERT embeddings amongst others. In addition, during this research study we have performed a plethora of different experiments in order to improve our systems performance.
Description: Πτυχιακή εργασία--Πανεπιστήμιο Μακεδονίας, Θεσσαλονίκη, 2023.
Η βιβλιοθήκη διαθέτει αντίτυπο της πτυχιακής μόνο σε ηλεκτρονική μορφή.
URI: http://dspace.lib.uom.gr/handle/2159/28676
Rights: Attribution-NonCommercial-NoDerivatives 4.0 Διεθνές
Appears in Collections:Τμήμα Εφαρμοσμένης Πληροφορικής (Π)

Files in This Item:
File Description SizeFormat 
ManoliChristinaPe2023.pdf4.21 MBAdobe PDFView/Open


This item is licensed under a Creative Commons License Creative Commons