Annotation manuelle de matchs de foot : Oh la la la ! l'accord inter-annotateurs ! et c'est le but !

Karen Fort (1) , Vincent Claveau (2)

1 LIPN - Laboratoire d'Informatique de Paris-Nord
2 TEXMEX - Multimedia content-based indexing

We present here an annotation campaign of commentaries of football matches in French. The annotation was done from a very heterogeneous text corpus of both match minutes and video commentary transcripts. We also show here how the intra- and inter-annotator agreements can be used efficiently during the whole campaign by proposing a definition suited to our type of task as well as emphasizing the impotance of using it appropriately. We also show how some clues collected through statistical analyses could be used to help correct the annotations. These statistical analyses are also used to assess the impact of the source modality (written or spoken) on the cost and quality of the annotation process.

Mots clés: annotation manuelle; accords inter-annotateurs

Domaines: Traitement du texte et du document

Karen Fort, Vincent Claveau. Annotation manuelle de matchs de foot : Oh la la la ! l'accord inter-annotateurs ! et c'est le but !. TALN - Traitement Automatique des Langues Naturelles, Jun 2012, Grenoble, France. pp.383-390. ⟨hal-00709181⟩