Sarah Shoilee is a PhD candidate at the Vrije Universiteit Amsterdam, where she conducts research at the intersection of knowledge engineering, information retrieval, and semantic web technologies. Her work focuses on designing methods that enable trustworthy knowledge discovery from complex, sparse, and heterogeneous data. Through her research, she aims to build knowledge systems that not only retrieve information, but also help people understand where knowledge comes from, how reliable it is, and how new insights can be discovered from imperfect data.
In her Research-in-Residence project at the KB, she aims to develop computational methods for linking historical events, people, and places in Dutch newspaper archives. Her research is structured around three main steps while addressing specific relevant challenges in these tasks:
- Event identification: Constructing a corpus of newspaper articles relevant to known historical events. Challenge: historical events are often referred to in different ways, with varying names and descriptions, making retrieval difficult.
- Named Entity Recognition: Applying named entity recognition (NER) to identify persons, places, and event mentions in those articles. Challenge: OCR errors, old spelling, and inconsistent punctuation make named entity recognition less reliable.
- Entity disambiguation: Disambiguating and linking these entities to connect people with the historical events they are associated with. Challenge: the same name may refer to different people, places, or events, and different spellings may refer to the same entity.
Ultimately, the aim of her project is to build a network of people associated with historical events in order to better understand how individuals are connected across different historical events and to shed light on the Dutch colonial past and the roles people played within it.
