Multi-News Parquet: Multi-Document News Summarization Dataset

A cleaned and updated version of the Multi-News dataset, converted to Parquet format for compatibility with the latest Hugging Face datasets library. The dataset was originally created by Alexander Fabbri et al. in 2019 and was last updated on August 22,我们发现一个错误，请关闭当前对话，重新开启一个新对话。

Use Cases

• Training abstractive summarization models based on multi-document news sources.
• Benchmarking extractive summarization algorithms on news article collections.
• Researching information fusion and redundancy detection across multiple documents.

Strengths

• Dataset is cleaned and updated from the original version.
• Converted to Parquet format for compatibility with modern data libraries.
• Last updated on 2025-08-22, suggesting recent maintenance.

Limitations

• Column-level documentation is absent; field semantics must be inferred after download.
• Row count, file size, and license information are unknown, which may limit suitability assessment.
• Description metadata is limited; actual data quality requires manual inspection after download.

Provenance

Source

Based on the original Multi-News dataset by Alexander Fabbri et al. (2019).

Collection Method

Cleaned, updated, and converted to Parquet format by Awesome075.

Time Range

Original dataset from 2019; parquet version updated in 2025.

Freshness

Last updated 2025-08-22 11:16:16.

Geography

null