Projection-based Coreference Resolution Using Deep Syntax

Abstract

The paper describes the system for coreference resolution in German and Russian, trained exclusively on coreference relations projected through a parallel corpus from English. The resolver operates on the level of deep syntax and makes use of multiple specialized models. It achieves 32 and 22 points in terms of CoNLL score for Russian and German, respectively. Analysis of the evaluation results show that the resolver for Russian is able to preserve 66% of the English resolver's quality in terms of CoNLL score. The system was submitted to the Closed track of the COR-BON 2017 Shared task.

Coreference Resolution: Toward End-to-End and Cross-Lingual Systems

“…Martins [98] applied projection on an English-Portuguese-Spanish parallel corpus, reporting strong results compared with other cross-lingual methods, such as bilingual word embeddings, bitext direct projection, or vanilla posterior regularization. For the same goal, Novák et al [99] reported improved cross-lingual results on the CORBON 2017 shared task [36], building German and Russian coreference resolvers based solely on parallel English data.…”

“…Cross-lingual knowledge transfer generally relies on either a common multi-lingual semantic space [49,96,97] or on parallel corpora, which enables projection of information from one language to another [98,99]. The use of multilingual word embeddings [49] as a multi-lingual semantic space enables the use of direct transfer learning between languages [100].…”

“…Several works have used projection-based techniques [102,103] for improving performance on less-resourced languages [98,99,104,105]. In projection approaches, learning instances originally in the source language are translated, using machine translation tools or parallel data, to the target language, and the corresponding labels are projected to the new learning instances in the target language.…”

A Fine-grained Large-scale Analysis of Coreference Projection

“…This approach has been used to create a coreference resolver by multiple authors, e.g. de Souza andOrȃsan (2011), Martins (2015), Wallin and Nugues (2017), and Novák et al (2017). However, since the present work employs the corpus-based approach on gold annotations of coreference, we offer more details on works of Postolache et al (2006) and Grishina andStede (2015, 2017).…”

Proceedings of the First Workshop on Computational Models of Reference, Anaphora and Coreference

“…This approach has been used to create a coreference resolver by multiple authors, e.g. de Souza and Orȃsan (2011), Martins (2015), Wallin and Nugues (2017), and Novák et al (2017). However, since the present work employs the corpus-based approach on gold annotations of coreference, we offer more details on works of and Stede (2015, 2017).…”