Joint entity and relation extraction via contrastive learning on knowledge-augmented graph embeddings
dc.contributor | Aalto-yliopisto | fi |
dc.contributor | Aalto University | en |
dc.contributor.advisor | Ji, Shaoxiong | |
dc.contributor.author | Gao, Ya | |
dc.contributor.school | Perustieteiden korkeakoulu | fi |
dc.contributor.supervisor | Marttinen, Pekka | |
dc.date.accessioned | 2023-01-29T18:09:39Z | |
dc.date.available | 2023-01-29T18:09:39Z | |
dc.date.issued | 2023-01-23 | |
dc.description.abstract | Entity Recognition (ER) and Relation Extraction (RE) are the two most critical tasks in information extraction. Rather than viewing them as two subtasks, recent studies are focusing on how to extract entities and relationships jointly, which is known as Joint Entity and Relation Extraction (JERE). However, in prior research, the interaction between entity recognition and relation extraction is not explicitly described. Besides, a lack of semantic and structural information leads to poor performance in extraction. These models also hard to handle the problem of entity overlapping, showing limitations of working in complex scenarios. To address these issues, in this work, we introduce the design of knowledge-augmented graph embeddings to enable existing models in capturing more information from the text and achieve a better understanding of the connections between entities and relations. In addition, we employ Contrastive Learning (CL) to encourage adaptive learning from external knowledge. Furthermore, we adopt a novel tagging scheme to transform this task into a triplet classification problem. Experimental results on three widely used datasets show a good performance of our model and illustrate the contributions of different blocks employed in the model. | en |
dc.format.extent | 50+2 | |
dc.identifier.uri | https://aaltodoc.aalto.fi/handle/123456789/119407 | |
dc.identifier.urn | URN:NBN:fi:aalto-202301291757 | |
dc.language.iso | en | en |
dc.programme | Master’s Programme in Computer, Communication and Information Sciences | fi |
dc.programme.major | Machine Learning, Data Science and Artificial Intelligence | fi |
dc.programme.mcode | SCI3044 | fi |
dc.subject.keyword | joint entity relation extraction | en |
dc.subject.keyword | graph neural networks | en |
dc.subject.keyword | contrastive learning | en |
dc.subject.keyword | knowledge graph | en |
dc.title | Joint entity and relation extraction via contrastive learning on knowledge-augmented graph embeddings | en |
dc.type | G2 Pro gradu, diplomityö | fi |
dc.type.ontasot | Master's thesis | en |
dc.type.ontasot | Diplomityö | fi |
local.aalto.electroniconly | yes | |
local.aalto.openaccess | no |