Schepers, Niek (2026) Very Important Propositions: Improving named entity recognition in Dutch historical documents using Text Region labels. Bachelor's Thesis, Artificial Intelligence.
|
Text
BachelorThesisAINiekSchepers.pdf Download (474kB) | Preview |
Abstract
To preserve the knowledge stored in historical documents, archives are often digitized. A part of this process is extracting and classifying named entities, important people and places within a text, through Named Entity Recognition (NER). The Collectie Overijssel, a Dutch archive, wants to digitize its collection of resolutions passed by the provincial state of Overijssel, but named entity recognition for Dutch historical documents presents many challenges. To over- come these challenges, text region labels that indicate the type of text (paragraph, header, etc.) are available. The goal of this paper was to determine whether adding these text region labels improves NER performance on Dutch historical documents. To determine this, an existing NER architecture for Dutch historical documents was adapted and fine-tuned, and several methods for incorporating text region labels were added. The resulting experiment showed that there is a statistically significant improvement in NER performance when adding text region labels for two out of five entities, capacity (t(18) = −2.97, p = .004) and location (t(17) = −2.71, p = .007). This shows that the additional context provided by text region labels improves NER performance on Dutch historical documents, indicating great potential for future NER research.
| Item Type: | Thesis (Bachelor's Thesis) |
|---|---|
| Supervisor name: | Dhali, M.A. and Schuijlenburg, K. van |
| Degree programme: | Artificial Intelligence |
| Thesis type: | Bachelor's Thesis |
| Language: | English |
| Date Deposited: | 05 Aug 2026 06:39 |
| Last Modified: | 05 Aug 2026 06:39 |
| URI: | https://fse.studenttheses.ub.rug.nl/id/eprint/37480 |
Actions (login required)
 |
View Item |