Unit:
Department of Informatics and TelecommunicationsΠληροφορική
Author:
CHALEPOUDIS GEORGIOS
Supervisors info:
Γουνόπουλος Δημήτριος, Καθηγητής, ΠΛΗΡΟΦΟΡΙΚΗΣ ΚΑΙ ΤΗΛΕΠΙΚΟΙΝΩΝΙΩΝ, Εθνικόν και Καποδιστριακόν Πανεπιστήμιον Αθηνών
Original Title:
Finding Exceptional Facts in Knowledge Graphs
Translated title:
Finding Exceptional Facts in Knowledge Graphs
Summary:
The wide use of the internet and its ability to introduce and amplify Fake News into the information stream demands to also make ways to detect these news, yet the immense amount of data makes it impossible to fact check every single thing and ways to figure out which facts take priority are needed. The study aims to explore the use of Knowledge Graphs in finding Exceptional Facts or in other words facts worth taking the time to fact check. In order to find these Exceptional Facts we make use of the algorithm outlined in the 2018 paper: Maverick Discovering Exceptional Facts from Knowledge Graphs.
The algorithm given an entity of interest will try to find exceptional facts about it. For the purposes of the study an interesting fact about an entity is when given a context that the entity is a part of it is in the minority of the entities for whom the fact applies. A scoring function is used to give a numerical value to the exceptionality . The algorithm works by constructing simple facts that are true for our entity of interest then finds contexts for which the entity of interest is exceptional given those facts, then it constructs more complex facts and repeats iteratively, until it has enough exceptional facts or it can’t find any more.
Results suggest that Knowledge Graphs can be used to find exceptional facts but are prone to showing a bias for facts that actually have no interest when interpreted to physical language and a need for better constructed Knowledge Graphs to avoid such pitfalls.
Main subject category:
Technology - Computer science
Keywords:
Knowledge Graph, Fake News, Exceptional Facts
