TY - JOUR
T1 - Low-resource AMR-to-Text Generation
T2 - A Study on Brazilian Portuguese
AU - Cabezudo, Marco Antonio Sobrevilla
AU - Pardo, Thiago Alexandre Salgueiro
N1 - Publisher Copyright:
© 2022 Sociedad Espanola para el Procesamiento del Lenguaje Natural. All rights reserved.
PY - 2022/3/1
Y1 - 2022/3/1
N2 - This work presents a study of how varied strategies for tackling lowresource AMR-to-text generation for three approaches are helpful in Brazilian Portuguese. Specifically, we explore the helpfulness of additional translated corpus, different granularity levels in input representation, and three preprocessing steps. Results show that translation is useful. However, it must be used in each approach differently. In addition, finer-grained representations as characters and subwords improve the performance and reduce the bias on the development set, and preprocessing steps are helpful in different contexts, being delexicalisation and preordering the most important ones.
AB - This work presents a study of how varied strategies for tackling lowresource AMR-to-text generation for three approaches are helpful in Brazilian Portuguese. Specifically, we explore the helpfulness of additional translated corpus, different granularity levels in input representation, and three preprocessing steps. Results show that translation is useful. However, it must be used in each approach differently. In addition, finer-grained representations as characters and subwords improve the performance and reduce the bias on the development set, and preprocessing steps are helpful in different contexts, being delexicalisation and preordering the most important ones.
KW - AMR-to-Text Generation
KW - Brazilian Portuguese
KW - Low-resource setting
UR - http://www.scopus.com/inward/record.url?scp=85128202739&partnerID=8YFLogxK
U2 - 10.26342/2022-68-6
DO - 10.26342/2022-68-6
M3 - Article
AN - SCOPUS:85128202739
SN - 1135-5948
VL - 68
SP - 85
EP - 97
JO - Procesamiento del Lenguaje Natural
JF - Procesamiento del Lenguaje Natural
ER -