Сравнение методов

	Многоязычное реферирование текста	Дообученная суммаризация текстов
Область	Глубокое обучение	Глубокое обучение
Семейство	Machine learning	Machine learning
Год появления	2020–2021	2019–2020
Автор метода	Multiple groups; popularized via mBART (Liu et al., 2020) and mT5 (Xue et al., 2021)	Lewis et al. (BART); Zhang et al. (PEGASUS); Raffel et al. (T5)
Тип	Seq2seq / encoder-decoder fine-tuning for summarization across languages	Fine-tuned sequence-to-sequence neural model
Основополагающий источник	Xue, L., Constant, N., Roberts, A., Kale, M., Al-Rfou, R., Siddhant, A., Barua, A., & Raffel, C. (2021). mT5: A Massively Multilingual Pre-Trained Text-to-Text Transformer. Proceedings of NAACL-HLT 2021, pp. 483–498. Association for Computational Linguistics.	Zhang, J., Zhao, Y., Saleh, M., & Liu, P. J. (2020). PEGASUS: Pre-training with Extracted Gap-sentences for Abstractive Summarization. Proceedings of the 37th International Conference on Machine Learning (ICML), 119, 11328–11339.
Другие названия	cross-lingual summarization, multilingual abstractive summarization, multilingual extractive summarization, multilingual seq2seq summarization	Fine-tuned summarization model, Abstractive summarization via fine-tuning, Seq2Seq fine-tuning for summarization, BART/T5/PEGASUS fine-tuning
Связанные	4	5
Сводка	Multilingual text summarization applies pre-trained multilingual encoder-decoder models — such as mT5 or mBART — to generate concise summaries of documents written in many languages, either within the same language (monolingual) or across languages (cross-lingual). Fine-tuning these models on multilingual summarization benchmarks like XL-Sum enables coverage of dozens of languages with a single model.	Fine-Tuned Text Summarization adapts a large pre-trained sequence-to-sequence model — such as BART, T5, or PEGASUS — to generate concise summaries of documents by training on domain-specific (document, summary) pairs. The approach yields substantially more fluent and faithful summaries than extractive or generic approaches by leveraging knowledge encoded in billions of pre-training tokens.