DACSA: A large-scale Dataset for Automatic summarization of Catalan and Spanish newspaper Articles

Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 5931 - 5943 July 10-15, 2022 ©2022 Association for Computational Linguistics DACSA: A large-scale Dataset for Automatic summarization of Catalan and Spanish newspaper Articles Encarna Segarra Vicent Ahuir§ Lluís-F. Hurtado José Ángel González VRAIN: Valencian Research Institute for Artificial Intelligence Universitat Politècnica de València, Spain {esegarra, lhurtado, jogomba2}@dsic.upv.es §viahes@eui.upv.es Abstract The application of supervised methods to au- tomatic summarization requires the availabil- ity of adequate corpora consisting of a set of document-summary pairs. As in most Natu- ral Language Processing tasks, the great ma- jority of available datasets for summarization are in English, making it difficult to develop automatic summarization models for other lan- guages. Although Spanish is gradually forming part of some recent summarization corpora, it is not the same for minority languages such as Catalan. In this work, we describe the con- struction of a corpus of Catalan and Spanish newspapers, the Dataset for Automatic sum- marization of Catalan and Spanish newspaper Articles (DACSA) corpus. It is a high-quality large-scale corpus that can be used to train sum- marization models for Catalan and Spanish. We have carried out an analysis of the corpus, both in terms of the style of the summaries and the difficulty of the summarization task. In particu- lar, we have used a set of well-known metrics in the summarization field in order to characterize the corpus. Additionally, we have evaluated the performance of some extractive and abstractive summarization systems on the DACSA corpus for benchmarking purposes. 1 Introduction Automatic summarization is one of the central prob- lems in Natural Language Processing (NLP). The development of automatic summarization systems is an important issue due to the great amount of information in different formats that is accessible on the web or in other repositories. It is necessary to develop techniques that help us to tackle that huge amount of information. For this reason, there is an increasing interest in the NLP community to develop techniques that allow the users to find, read, understand, or process the documents. In this context, automatic summarization can be an impor- tant aid because it provides a condensed version of documents that reduce the time to explore or analyze them. Access to large-scale high-quality data is an es- sential prerequisite for making substantial progress in summarization. The application of supervised methods to automatic summarization, as those based on Neural Networks, requires the availabil- ity of adequate corpora consisting of document- summary pairs. The construction of large-scale and high-quality corpora for learning neural summariza- tion models is not an easy task. It is necessary a great human effort to generate thousands of man- ual summaries, or to design new approaches to ob- tain these summaries in a semiautomatic way. The first important resource for learning corpus-based summarization models was the CNN/DailyMail summarization corpus (Hermann et al., 2015), orig- inally constructed for the task of passage-based question answering and adapted to the document …

Content truncated to 500 words