VLSP 2022 Abmusu Task Dataset: A Resource for Vietnamese Abstractive Multi-Document Summarization

• DOI: 10.1142/s2717554523500030
• Publication Date: 2023-05-17T07:18:47Z
• Authors: Quoc-An Nguyen; Duy-Cat Can; Hoang-Quynh Le; Mai-Vu Tran
• Keywords: Vietnamese; Benchmark (surveying); Multi-document summarization

The performance of automatic summarization systems has improved significantly with the development supervised approaches. However, in Vietnamese abstractive multi-document task, available datasets are insufficient for training model. With this motivation, we contribute a new gold standard dataset, named Abmusu. Following collecting and clustering articles, have built hierarchical annotation process to generate summaries, three roles: annotator, supervisor, curator. As result, dataset contains 600 news clusters formed from 1839 articles corresponding human-generated summaries. To best our knowledge, Abmusu is biggest that freely research. Moreover, summaries more concise, making it challenging train models. We also used various baselines benchmark dataset.