Two New Datasets for Italian-Language Abstractive Text Summarization

Text summarization aims to produce a short summary containing relevant parts from a given text. Due to the lack of data for abstractive summarization on low-resource languages such as Italian, we propose two new original datasets collected from two Italian news websites with multi-sentence summaries and corresponding articles, and from a dataset obtained by machine translation of a Spanish summarization dataset. These two datasets are currently the only two available in Italian for this task. To evaluate the quality of these two datasets, we used them to train a T5-base model and an mBART model, obtaining good results with both. To better evaluate the results obtained, we also compared the same models trained on automatically translated datasets, and the resulting summaries in the same training language, with the automatically translated summaries, which demonstrated the superiority of the models obtained from the proposed datasets.

PDF Abstract
Task Dataset Model Metric Name Metric Value Global Rank Benchmark
Abstractive Text Summarization Abstractive Text Summarization from Fanpage mBART ROUGE-1 36.50 # 2
Abstractive Text Summarization Abstractive Text Summarization from Fanpage IT5 ROUGE-1 33.83 # 6
Abstractive Text Summarization Abstractive Text Summarization from Il Post IT5 ROUGE-1 33.78 # 5
Abstractive Text Summarization Abstractive Text Summarization from Il Post Pegasus-CNN/DM (eng-it translation) ROUGE-1 23.96 # 7
Abstractive Text Summarization Abstractive Text Summarization from Il Post Pegasus-XSum (eng-it translation) ROUGE-1 21.03 # 8
Abstractive Text Summarization Abstractive Text Summarization from Il Post mBART ROUGE-1 38.91 # 1
Abstractive Text Summarization MLSum-it IT5 rouge1 19.29 # 2
Abstractive Text Summarization MLSum-it Pegasus-CNN/DM (eng-it translation) rouge1 16.97 # 3
Abstractive Text Summarization MLSum-it Pegasus-XSum (eng-it translation) rouge1 15.17 # 4
Abstractive Text Summarization MLSum-it mBART rouge1 19.35 # 1

Methods