Fiche du document
29 juin 2015
- handle: 10670/1.2inrmw
- hal: hal-01348549
info:eu-repo/semantics/OpenAccess
Sujets proches
Frenchmen (French people)Citer ce document
Agnès Tutin et al., « Annotation of multiword expressions in French », HAL-SHS : linguistique, ID : 10670/1.2inrmw
Métriques
Partage / Export
Résumé
This paper presents an experiment of annotation of MWEs in French. The corpus used is made of several genres (news, novel, scientific report, film subtitles) and includes a rich annotation scheme including several kinds of MWEs from collocations to routines and full phrasemes. The annotation is performed semi-automatically with finite-state transducers. The inter-annotator agreement score shows that the annotation is quite consistent but the difficulty of the task relies heavily on the textual genre: literary texts are harder to annotate than scientific reports. Besides, two types of categories are difficult to differentiate, collocations and full phrasemes.