| Author: | Dong, Junnan |
| Title: | Integrating entity attributes for detecting errors in knowledge graphs |
| Advisors: | Huang, Xiao (COMP) |
| Degree: | M.Sc. |
| Year: | 2021 |
| Subject: | Semantic networks (Information theory) Knowledge management Artificial intelligence Hong Kong Polytechnic University -- Dissertations |
| Department: | Department of Computing |
| Pages: | vii, 46 pages : illustrations |
| Language: | English |
| Abstract: | Knowledge graphs (KGs) are gaining attention due to their expressivity and flexibility in storing and handling complex data. As essential supporters for knowledge-driven artificial intelligent systems, most KGs are automatically constructed from online corpus or text information. However, erroneous facts are inevitably injected into KGs due to original noise in these sources and imperfect extraction methods. To single out the correctness of facts (i.e., triples) inside a Knowledge Graph, a feasible way is to map the KG into vector space by leveraging the semantics of triples. Although many representation learning methods have been designed for KGs, such kind of methods are not suitable for error detection task. Specifically, existing representation learning methods are usually established based on a strong assumption that all observed triples are originally right so that may overfit some potential noises and fail the task to detect these errors. To this end, we propose a novel knowledge graph error detection framework, which integrating entity attributes to automatically detect errors in KGs. Specifically, we treat the original KG without attributes information as pure graph, and build an attribute graph based on these side-information and tread it as a congruent view of target KG. The confidence score of each triple is calculated by considering: 1) self-contradictory within the triple; 2) local-global consistency in graph structure; and 3) structure-attribute homogeneity. Performances on two real-world KGs demonstrate that our error detection method outperforms the current state-of-the-art KG error detection algorithms evaluated by two different evaluation metrics, with comparable or even better running efficiency. |
| Rights: | All rights reserved |
| Access: | restricted access |
Files in This Item:
| File | Description | Size | Format | |
|---|---|---|---|---|
| 5818.pdf | For All Users (off-campus access for PolyU Staff & Students only) | 707.63 kB | Adobe PDF | View/Open |
Copyright Undertaking
As a bona fide Library user, I declare that:
- I will abide by the rules and legal ordinances governing copyright regarding the use of the Database.
- I will use the Database for the purpose of my research or private study only and not for circulation or further reproduction or any other purpose.
- I agree to indemnify and hold the University harmless from and against any loss, damage, cost, liability or expenses arising from copyright infringement or unauthorized usage.
By downloading any item(s) listed above, you acknowledge that you have read and understood the copyright undertaking as stated above, and agree to be bound by all of its terms.
Please use this identifier to cite or link to this item:
https://theses.lib.polyu.edu.hk/handle/200/11370