An Investigation of Transformer-based Event Type Classification Through the Linguistic Lens
An Investigation of Transformer-based Event Type Classification Through the Linguistic Lens
View/Type of Scholarly Publication
DissertationDate of Exam
23.06.2026Date of Publication
13.07.2026Advisor
Kranich, SvenjaCo-Referee
Schröder, BernhardMetadata
Show full item record
Citable Links 
Abstract
Natural language processing implemented with the help of transformer models has proven to be a useful building block for the automatic classification of texts according to their event types. Transformer models learn features for classification automatically during training. The process of feature selection is opaque to the researcher.
This thesis analyzes the combination of transformer models and a set of manually chosen linguistically informed features as a way of approaching the task of event type classification. Models that use these linguistic features for classification are contrasted with models that do not. Advantages and disadvantages of the approach are evaluated and discussed through a series of experiments.
This thesis analyzes the combination of transformer models and a set of manually chosen linguistically informed features as a way of approaching the task of event type classification. Models that use these linguistic features for classification are contrasted with models that do not. Advantages and disadvantages of the approach are evaluated and discussed through a series of experiments.
Subjects
Textklassifikation, Künstliche Intelligenz, Eventdetektion
Classification (DDC)
400 Sprache, LinguistikNindel, Theresa: An Investigation of Transformer-based Event Type Classification Through the Linguistic Lens. - Bonn, 2026. - Dissertation, Rheinische Friedrich-Wilhelms-Universität Bonn.
Online-Ausgabe in bonndoc: https://nbn-resolving.org/urn:nbn:de:hbz:5-91210
Online-Ausgabe in bonndoc: https://nbn-resolving.org/urn:nbn:de:hbz:5-91210
@phdthesis{handle:20.500.11811/14277,
urn: https://nbn-resolving.org/urn:nbn:de:hbz:5-91210,
author = {{Theresa Nindel}},
title = {An Investigation of Transformer-based Event Type Classification Through the Linguistic Lens},
school = {Rheinische Friedrich-Wilhelms-Universität Bonn},
year = 2026,
month = jul,
note = {Natural language processing implemented with the help of transformer models has proven to be a useful building block for the automatic classification of texts according to their event types. Transformer models learn features for classification automatically during training. The process of feature selection is opaque to the researcher.
This thesis analyzes the combination of transformer models and a set of manually chosen linguistically informed features as a way of approaching the task of event type classification. Models that use these linguistic features for classification are contrasted with models that do not. Advantages and disadvantages of the approach are evaluated and discussed through a series of experiments.},
url = {https://hdl.handle.net/20.500.11811/14277}
}
urn: https://nbn-resolving.org/urn:nbn:de:hbz:5-91210,
author = {{Theresa Nindel}},
title = {An Investigation of Transformer-based Event Type Classification Through the Linguistic Lens},
school = {Rheinische Friedrich-Wilhelms-Universität Bonn},
year = 2026,
month = jul,
note = {Natural language processing implemented with the help of transformer models has proven to be a useful building block for the automatic classification of texts according to their event types. Transformer models learn features for classification automatically during training. The process of feature selection is opaque to the researcher.
This thesis analyzes the combination of transformer models and a set of manually chosen linguistically informed features as a way of approaching the task of event type classification. Models that use these linguistic features for classification are contrasted with models that do not. Advantages and disadvantages of the approach are evaluated and discussed through a series of experiments.},
url = {https://hdl.handle.net/20.500.11811/14277}
}
The following license files are associated with this item:
