🧑‍🔬 Maria Antònia Martí

Maria Antònia Martí works on computational linguistics, with a focus on building and annotating language resources. Their papers here deal with coreference resolution and with semantic role labeling across languages. A recurring question is how annotation schemes should handle cases that do not fit clean categories. The work draws on corpora in English and Catalan, among other languages.

The most cited of these papers is The CoNLL-2009 Shared Task, an overview of a shared evaluation on joint dependency parsing and semantic role labelling. It carried an earlier English-only task over to Catalan, Chinese, Czech, German, Japanese and Spanish, using a common dependency format. It also offered closed and open challenges and added out-of-domain test sets for some languages. The next most cited, A Typology of Near-Identity Relations for Coreference, argues that coreference is better seen as a continuum than as a yes-or-no relation. It proposes near-identity as a middle category, for example a person versus a photograph of that person, and builds the NIDENT typology from examples found in real text.

The latest work, Annotating Near-Identity from Coreference Disagreements from 2012, tests that idea in practice. It adds near-identity labels to portions of the English NP4E corpus and the Catalan AnCora-CA corpus. Annotators rarely chose these labels when asked directly, so the paper also tries an indirect route that uses disagreement among annotators as the signal. Martí's papers are associated with Universitat de Barcelona. The coreference studies were written with Marta Recasens, and the CoNLL overview with a broad multilingual team including Jan Hajič and Lluís Màrquez.

Maria Antònia Martí has 3 papers in our library, in 2009–2012, most of them on Coreference resolution, Semantic role labeling, Multilingual & cross-lingual NLP; often with Marta Recasens, Adam Meyers, Constantin Orasan.