| Texto completo | |
| Autor(es): |
Zani Jorge, Germano Antonio
;
Bezerra, Davi Alves
;
Castell, Clarissa
;
Salgueiro Pardo, Thiago Alexandre
Número total de Autores: 4
|
| Tipo de documento: | Artigo Científico |
| Fonte: | ADVANCES IN ARTIFICIAL INTELLIGENCE-IBERAMIA 2024; v. 15277, p. 5-pg., 2025-01-01. |
| Resumo | |
Automatic Text Summarization is an important task for managing large volumes of information. Summarization methods include extractive, which select text passages directly from the source(s), and abstractive ones, which generate new sentences from the original content [10]. Some current state-of-the-art extractive single-document summarization methods like PreSumm [6], HiStruct+ [12], and MemSum [4] have shown success in English corpora but are under-explored for other languages. In this paper, we test such methods for Brazilian Portuguese, considered a low-resource language. Besides the traditional ROUGE informativeness evaluation metric [5], we also use the more recent BLANC metric [13], which assesses summary fluency and coherence. We show that the methods also perform well for Portuguese. (AU) | |
| Processo FAPESP: | 19/07665-4 - Centro de Inteligência Artificial |
| Beneficiário: | Fabio Gagliardi Cozman |
| Modalidade de apoio: | Auxílio à Pesquisa - Programa eScience e Data Science - Centros de Pesquisa Aplicada |