Analyse, visualisation et identification automatique des chaînes de coréférences : des questions interdépendantes ?

Abstract

Une chaîne de coréférences est une structure qui regroupe un ensemble d'expressions référentielles (ou mentions, ou maillons) désignant toutes la même entité extralinguistique. Chaque maillon peut être enrichi par des annotations linguistiques, de même que les relations reliant certains maillons. En conséquence, il est difficile d'appréhender une telle structure et d'en tirer directement des analyses. Nous présentons des repères méthodologiques importants pour favoriser l'exploitation d'un corpus annoté en chaînes, tout en précisant les liens opérant entre linguistique théorique, linguistique de corpus outillée et traitement automatique.

Computational Models of Anaphora

“…The discussions about the specification of the coreference task in muc and ace (van Deemter & Kibble 2000) eventually led to proposals for the annotation of anaphoric information (Passonneau 1997, Poesio et al 1999) that were more directly based on the linguistic approach to anaphora discussed in Section 2.1. Most of the corpora developed since have adopted a similar approach (Poesio 2004, Hinrichs et al 2005, Hendrickx et al 2008, Poesio & Artstein 2008, Nedoluzhko et al 2009, Pradhan et al 2012, Ogrodniczuk et al 2015, Landragin 2016, Zeldes 2020. In particular, the creation of OntoNotes (Pradhan et al 2012) and the shared tasks based on OntoNotes and other data sets of this type , Pradhan et al 2012) led to a move away from the modeling of coreference in the sense of muc and ace and toward anaphora resolution as traditionally conceived in linguistics and psychology.…”

Exploring Social Sciences Archives with Explainable Document Linkage through Question Generation

“…There are also end-to-end coreference resolution systems for French, such as DeCOFre (Grobol, 2020) and coFR (Wilkens et al, 2020). DeCOFre 7 is trained primarily on spontaneous spoken language (ANCOR corpus, (Muzerelle et al, 2013)), while coFR 8 is trained on both spoken (ANCOR corpus) and written language (Democrat corpus, (Landragin, 2016)). For this study, we use coFR, as it is better suited for our corpus (i.e., archives and documents).…”

Proceedings of the 7th Joint SIGHUM Workshop on Computational Linguistics for Cultural Heritage, Social Sciences, Humanities and Literature

“…There are also end-to-end coreference resolution systems for French, such as DeCOFre (Grobol, 2020) and coFR (Wilkens et al, 2020). DeCOFre 7 is trained primarily on spontaneous spoken language (ANCOR corpus, (Muzerelle et al, 2013)), while coFR 8 is trained on both spoken (ANCOR corpus) and written language (Democrat corpus, (Landragin, 2016)). For this study, we use coFR, as it is better suited for our corpus (i.e., archives and documents).…”