Missing Link Prediction in Art Knowledge Graph using Representation Learning
Swapnil S. Mahure1, Anish R. Khobragade2
1Swapnil S. Mahure, College of Engineering, COEP Technological University Pune (Maharashtra), India.
2Anish R. Khobragade, College of Engineering, COEP Technological University Pune (Maharashtra), India.
Manuscript received on 18 August 2022 | Revised Manuscript received on 03 September 2022 | Manuscript Accepted on 15 April 2024 | Manuscript published on 30 April 2024 | PP: 30-33 | Volume-13 Issue-5, April 2024 | Retrieval Number: 100.1/ijitee.J926409111022 | DOI: 10.35940/ijitee.J9264.13050424
Open Access | Editorial and Publishing Policies | Cite | Zenodo | OJS | Indexing and Abstracting
© The Authors. Blue Eyes Intelligence Engineering and Sciences Publication (BEIESP). This is an open access article under the CC-BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/)
Abstract: Knowledge graphs are an essential and evolving field in the Artificial Intelligence domain, with multiple applications, including question answering, information retrieval, information recommendation, and Natural language processing. A knowledge graph has one significant limitation: incompleteness. This is because real-world data are dynamic and continue to evolve. The incompleteness of the Knowledge graph can be overcome or minimised by using representation learning models. Several models are classified based on translation distance, semantic information, and NN (Neural Network) models. Earlier, various embedding models were tested on mainly two well-known datasets: WN18RR and FB15k-237. In this paper, a new dataset, namely ArtGraph, has been utilised for link prediction using representation learning models to enhance the application of ArtGraph in various domains. Experimental results show that ConvKB outperformed the other models in the link prediction task.
Keywords: KG Embeddings, Artwork, Link Prediction, Neural Network
Scope of the Article: Neural Information Processing