Material Philology Meets Digital Onomastic Lexicography: The NordiCon Database of Medieval Nordic Personal Names in Continental Sources

Proceedings of the 12th Conference on Language Resources and Evaluation (LREC 2020), pages 80–89 Marseille, 11–16 May 2020 c⃝ European Language Resources Association (ELRA), licensed under CC-BY-NC 80 French coreference for spoken and written language Rodrigo Wilkens1, Bruno Oberle1, Fr´ed´eric Landragin2, Amalia Todirascu1 1 LiLPa, 2 LaTTICe-CNRS 1 University of Strasbourg, 2 ´Ecole Normale Sup´erieure & Universit´e Paris 3-Sorbonne Nouvelle rswilkens@gmail.com, b.oberle@zoho.eu, frederic.landragin@ens.fr, todiras@unistra.fr Abstract Coreference resolution aims at identifying and grouping all mentions referring to the same entity. In French, most systems run different setups, making their comparison difficult. In this paper, we present an extensive comparison of several coreference resolution systems for French. The systems have been trained on two corpora (ANCOR for spoken language and Democrat for written language) annotated with coreference chains, and augmented with syntactic and semantic information. The models are compared with different configurations (e.g. with and without singletons). In addition, we evaluate mention detection and coreference resolution apart. We present a full-stack model that outperforms other approaches. This model allows us to study the impact of mention detection errors on coreference resolution. Our analysis shows that mention detection can be improved by focusing on boundary identification while advances in the pronoun-noun relation detection can help the coreference task. Another contribution of this work is the first end-to-end neural French coreference resolution model trained on Democrat (written texts), which compares to the state-of-the-art systems for oral French. Keywords: Coreference, Anaphora, Automatic Coreference Resolution, French, Evaluation 1. Introduction The coreference resolution task aims at identifying all men- tions that refer to the same extra-linguistic entity (Pradhan et al., 2012; Recasens and Pradhan, 2016). It may be di- vided into two subtasks: mention detection and corefer- ence resolution. For example, in Text (1), Justin Trudeau, Canada’s prime minister, his and he are mentions referring to the same entity (the person named Justin Trudeau). They form a coreference chain. (1) [Justin Trudeau] has won a second term after the country’s federal election, but [his] narrow victory means the [Canada’s prime minister] will lead a mi- nority government that will be forced to depend on other parties to govern (adapted from The Guardian, 11/29/2019). This task is difficult because of the linguistic and extra- linguistic knowledge required to link various mentions into a single coreference chain (Mitkov, 2002). For example, to find the antecedent of he, one must know that Justin Trudeau is a person (otherwise the pronoun would have been it) and that Justin is a male first name. Syntac- tic knowledge may also prove useful, since both Justin Trudeau and he are subjects (syntactic parallelism (Poesio et al., 2016; Ng, 2017)). Semantic knowledge, on the other hand, gives the information that prime minister is a pub- lic title held by a person. Finally, world knowledge may help to find the correct referent, since Justin Trudeau is the actual prime minister of Canada. Approaches that address this task automatically may be classified in different ways. For instance, Ng (2017), who focuses on methods used to link mentions …

Content truncated to 500 words