dc.rights.license | CC-BY-NC-ND | |
dc.contributor.advisor | Fowlie, M. | |
dc.contributor.author | Hanneman, K.A. | |
dc.date.accessioned | 2020-05-08T18:00:11Z | |
dc.date.available | 2020-05-08T18:00:11Z | |
dc.date.issued | 2020 | |
dc.identifier.uri | https://studenttheses.uu.nl/handle/20.500.12932/35794 | |
dc.description.abstract | Although in relation extraction, dependency based models are currently outclassed by a variety of models using different techniques, these dependency models have shown promising results. Here syntactic dependency trees are predominantly employed. However, consequently it is uncertain what performance semantic dependency graphs offer in comparison. Therefore we propose a modified graph convolutional network for relation extraction to work with semantic dependency graphs instead of syntactic dependency trees. The performance of this model is tested on the TACRED dataset, where for each entry in this dataset semantic dependency graphs are generated with a state-of-the-art model. Certain sentences within the set were not included for this study, as these could not be parsed. The results then of this model show increased performance when presented less training data, and equal performance when the amount of data managed to increase. | |
dc.description.sponsorship | Utrecht University | |
dc.format.extent | 401860 | |
dc.format.mimetype | application/pdf | |
dc.language.iso | en | |
dc.title | Semantic Dependency Graph Convolution for Relation Extraction | |
dc.type.content | Bachelor Thesis | |
dc.rights.accessrights | Open Access | |
dc.subject.keywords | relation extraction, nlp, semantic dependency graph, machine learning, ai, tacred, graph convolutional network | |
dc.subject.courseuu | Kunstmatige Intelligentie | |