Education

2019 - 2022

In the current era of massive digitization of historical sources, the automatic extraction of events is a crucial step when dealing with historical texts. Event processing is an active area of research in the Natural Language Processing community, but resources and systems are mainly developed for processing contemporary texts. In this context, this thesis aims at automatically extracting events from historical documents. It proposes multidisciplinary exchanges in order to adapt recent ontologies to historical research purposes. Beyond the specific needs of the digital humanities, OCR’d historical documents dating back more than a century pose major problems in NLP (diachrony, quality, domain adaptation). Domain adaptation techniques combining specialised architectures and pre-⁠processing steps make it possible to reduce the impact of these difficulties. Based on a recent paradigm that casts tasks as a question answering problem, an event extraction pipeline is proposed: from the extraction of a trigger word in a sentence to the representation of more than a century of events in the form of graphs.

2017 - 2019

2016