ViMs: a high-quality Vietnamese dataset for abstractive multi-document summarization
- PDF / 2,011,450 Bytes
- 28 Pages / 439.37 x 666.142 pts Page_size
- 92 Downloads / 190 Views
ViMs: a high-quality Vietnamese dataset for abstractive multi-document summarization Nhi-Thao Tran1 • Minh-Quoc Nghiem1 • Nhung T. H. Nguyen1 • Ngan Luu-Thuy Nguyen2 Nam Van Chi1 • Dien Dinh1
•
Ó Springer Nature B.V. 2020
Abstract Automatic text summarization is important in this era due to the exponential growth of documents available on the Internet. In the Vietnamese language, VietnameseMDS is the only publicly available dataset for this task. Although the dataset has 199 clusters, there are only three documents in each cluster, which is small compared to typical datasets in English. This motivates us to construct ViMs—a big and high-quality Vietnamese dataset for abstractive multi-document summarization. To that end, we recruited 29 annotators and enhanced MDSWriter—an open-source annotation tool, to support the annotators in creating gold standard summaries. As a result, ViMs has 600 summaries corresponding to 300 clusters of 1,945 documents. We have verified the reliability of our dataset by using a variety of metrics including conventional Cohen’s j, relaxed Cohen’s j—a new metric that we propose to make it more suitable for abstractive summarization, and
& Nhi-Thao Tran [email protected] Minh-Quoc Nghiem [email protected] Nhung T. H. Nguyen [email protected] Ngan Luu-Thuy Nguyen [email protected] Nam Van Chi [email protected] Dien Dinh [email protected] 1
Faculty of Information Technology, HCMC University of Science, Ho Chi Minh City, Vietnam
2
Faculty of Computer Science, HCMC University of Information Technology, Ho Chi Minh City, Vietnam
123
Lang Resources & Evaluation
ROUGE scores. A relaxed j score of 0.55 indicate that ViMs could attain moderate agreement between annotators. Meanwhile, ROUGE scores are 0.729 of ROUGE-1, 0.507 of ROUGE-2 and 0.524 of ROUGE-SU4. We have further evaluated ViMs by using three different summarization systems: TextRank, CFVi and MUSEEC. Their performances are 0.628, 0.711 and 0.732 of ROUGE-1, respectively. These results show that the ViMs dataset is suitable for both training and evaluating multidocument summarization systems. We have made the dataset and evaluation results of this work publicly available for research community. It is noted that unlike previous work that only published the final summarization dataset, we also publish intermediate annotation results, which can be used in other NLP problems such as sentence classification. Keywords Abstractive summarization Multi-document summarization Vietnamese dataset Automatic summarization
1 Introduction In recent years, there has been an increasing interest in automatic text summarization due to the exponential growth of documents available on the Internet. Automatic text summarization is the process of using softwares to create a concise and fluent summary that contains major points of a document or a set of documents. Such automatic systems can save us time in digesting tons of information coming from the Internet. Based on the input, text summarization can be categori
Data Loading...