À la recherche des haïdouks. Annoter les entités spatiales dans des romans roumains du xixe siècle

This paper presents the annotation of spatial entities in a corpus of Romanian novels written mainly in the 19th century and featuring as their main character a figure very well-known in the Balkans, the Hajduk outlaw. Because of their nomadic lifestyle, Hajduks were experiencing a unique relation to space, which was very different from that of the majority of Romanians, who lived in villages at the time. To verify if this different vision reverberates in novels, the aim was to collect all nominal and adverbial phrases referring to space, thus going largely beyond toponyms, that are usually marked-up and examined in digital humanities and natural language processing projects. After a presentation of the main differences between our approach and those prevailing in older or more recent studies in spatial literary analyses, we focus on the elaboration of the annotation scheme and comment upon the results of the inter-annotator agreement, as measured on a set of ten fully-annotated samples, of a thousand words each, from our novels.