CSTNews - A Discourse-Annotated Corpus for Single and Multi-Document Summarization of News Texts in Brazilian Portuguese
Paula Christina Figueira Cardoso, Erick Galani Maziero, Maria Giuseppina Lucia, Ramirez Jorge, Ariani Di Felippo, Lúcia Helena Machado Rino, Maria das Graças, V. Nunes, Thiago Alexandre Salgueiro Pardo, Rodovia Washington Luís · 2011
Summary. This paper introduces CSTNews, a discourse-annotated corpus for fostering research on single and multi-document summarization. The corpus comprises 50 clusters of news texts in Brazilian Portuguese and some related material, which includes a set of single-document manual summaries and a set of multi-document manual and automatic summaries. The texts are annotated in different ways for discourse organization, following both the Rhetorical Structure Theory and Cross-document Structure Theory. The corpus is a result delivered within the context of the SUCINTO Project, which aims at investigating summarization strategies and developing tools and resources for that purpose. The design of the discourse annotation tasks and the decisions that have been taken during the annotation process are detailed in this paper.