A Data-driven Approach to Named Entity Recognition for Early Modern French

Proceedings of the 29th International Conference on Computational Linguistics, pages 3722–3730 October 12–17, 2022. 3722 A Data-driven Approach to Named Entity Recognition for Early Modern French Pedro Ortiz Suarez Data and Web Science Group, University of Mannheim B 6, 26 68159 Mannheim, Germany pedro.ortiz@uni-mannheim.de Simon Gabay Université de Genève Rue du Général-Dufour 24, 1211 Genève, Switzerland simon.gabay@unige.ch Abstract Named entity recognition has become an in- creasingly useful tool for digital humanities research, specially when it comes to histor- ical texts. However, historical texts pose a wide range of challenges to both named entity recognition and natural language processing in general that are still difficult to address even with modern neural methods. In this article we focus in named entity recognition for histor- ical French, and in particular for Early Mod- ern French (16th-18th c.), i.e. Ancien Régime French. However, instead of developing a spe- cialised architecture to tackle the particularities of this state of language, we opt for a data- driven approach by developing a new corpus with fine-grained entity annotation, covering three centuries of literature corresponding to the early modern period; we try to annotate as much data as possible producing a corpus that is many times bigger than the most popular NER evaluation corpora for both Contempo- rary English and French. We then fine-tune existing state-of-the-art architectures for Early Modern and Contemporary French, obtaining results that are on par with those of the current state-of-the-art NER systems for Contemporary English. Both the corpus and the fine-tuned models are released. 1 Introduction Named entity recognition (NER) is an extensively studied task in natural language processing (NLP) that consists in identifying and classifying named entities mentions in unstructured text. These named entities often are real-world objects such as a per- son, a location, an organisation name or even a product. NER has been an important task in nat- ural language processing for some time now. It was the focus of the MUC conferences and associ- ated shared tasks (Marsh and Perzanowski, 1998), and later that of the CoNLL 2003 and the ACE shared tasks (Tjong Kim Sang and De Meulder, 2003; Doddington et al., 2004). NER has quickly established itself as a pillar of the new methods of reading texts promoted by the digital humanities (DH), based on the analy- sis of large sets of literary or historical data via computational methods (Moretti, 2005). These sources being not only contemporary, the need for tools dealing with medieval or early modern states of language is now increasing. NER interests re- searchers in DH for numerous reasons since the application can be quite broad, from genealogy or history studies for which finding mentions of persons and places in texts is very useful; to appli- cations in digital literature where researchers can use NER to highlight the path of different charac- ters in a book or in a series of publications. Both the research in NER and DH can benefit from one another as it has already been suggested particular properties of literature …

Content truncated to 500 words