Multi-News, consists of news articles and human-written summaries of these articles from the site newser.com. Each summary is professionally written by editors and includes links to the original articles cited. Source: Multi-News: a Large-Scale Multi-Document Summarization Dataset and Abstractive Hierarchical Model Image Source: https://arxiv.org/pdf/1906.01749.pdf

WikiSum is a dataset based on English Wikipedia and suitable for a task of multi-document abstractive summarization. In each instance, the input is comprised of a Wikipedia topic (title of article) and a collection of non-Wikipedia reference documents, and the target is the Wikipedia article text. The dataset is restricted to the articles with at least one crawlable citation. The official split divides the articles roughly into 80/10/10 for train/development/test subsets, resulting in 1865750, 233252, and 232998 examples respectively. Source: Generating Wikipedia by Summarizing Long Sequences Image Source: https://arxiv.org/pdf/1801.10198.pdf

The WCEP dataset for multi-document summarization (MDS) consists of short, human-written summaries about news events, obtained from the Wikipedia Current Events Portal (WCEP), each paired with a cluster of news articles associated with an event. These articles consist of sources cited by editors on WCEP, and are extended with articles automatically obtained from the Common Crawl News dataset. Source: WCEP

The DUC2004 dataset is a dataset for document summarization. Is designed and used for testing only. It consists of 500 news articles, each paired with four human written summaries. Specifically it consists of 50 clusters of Text REtrieval Conference (TREC) documents, from the following collections: AP newswire, 1998-2000; New York Times newswire, 1998-2000; Xinhua News Agency (English version), 1996-2000. Each cluster contained on average 10 documents. Source: Discrete Optimization for Unsupervised Sentence Summarization with Word-Level Extraction Image Source: https://duc.nist.gov/duc2004/

5,519 query-based summaries, each associated with an average of 6 input documents selected from an index of 355M documents from Common Crawl. Source: AQuaMuSe: Automatically Generating Datasets for Query-Based Multi-Document Summarization

Multi-XScience is a large-scale dataset for multi-document summarization of scientific articles. It has 30,369, 5,066 and 5,093 samples for the train, validation and test split respectively. The average document length is 778.08 words and the average summary length is 116.44 words. Source: https://github.com/yaolu/Multi-XScience

OPOSUM is a dataset for the training and evaluation of Opinion Summarization models which contains Amazon reviews from six product domains: Laptop Bags, Bluetooth Headsets, Boots, Keyboards, Televisions, and Vacuums. The six training collections were created by downsampling from the Amazon Product Dataset introduced in McAuley et al. (2015) and contain reviews and their respective ratings. A subset of the dataset has been manually annotated, specifically, for each domain, 10 different products were uniformly sampled (across ratings) with 10 reviews each, amounting to a total of 600 reviews, to be used only for development (300) and testing (300). Source: Summarizing Opinions: Aspect Extraction Meets Sentiment Prediction and They Are Both Weakly Supervised

Maternal and Infant (MATINF) Dataset is a large-scale dataset jointly labeled for classification, question answering and summarization in the domain of maternity and baby caring in Chinese. An entry in the dataset includes four fields: question (Q), description (D), class (C) and answer (A). Nearly two million question-answer pairs are collected with fine-grained human-labeled classes from a large Chinese maternity and baby caring QA site. Authors conduct both automatic and manual data cleansing and remove: (1) classes with insufficient samples; (2) entries in which the length of the description filed is less than the length of the question field; (3) data with any field longer than 256 characters; (4) human-spotted ill-formed data. After the data cleansing, MATINF is constructed with the remaining 1.07 million entries Source: MATINF: A Jointly Labeled Large-Scale Dataset for Classification, Question Answering and Summarization

MS^2 (Multi-Document Summarization of Medical Studies) is a dataset of over 470k documents and 20k summaries derived from the scientific literature. This dataset facilitates the development of systems that can assess and aggregate contradictory evidence across multiple studies, and is one of the first large-scale, publicly available multi-document summarization dataset in the biomedical domain.

Wikipedia Generation is a dataset for article generation from Wikipedia from references at the end of Wikipedia page and the top 10 search results for the Wikipedia topic.

Healthline is a nutrition related dataset for multi-document summarization, using scientific studies.

OpenAsp Dataset OpenAsp is an Open Aspect-based Multi-Document Summarization dataset derived from DUC and MultiNews summarization datasets. Antarctica ## Dataset Access To generate OpenAsp, you require access to the DUC dataset which OpenAsp is derived from. ### Steps: 1. Grant access to DUC dataset by following NIST instructions here. * you should receive two user-password pairs (for DUC01-02 and DUC06-07) * you should receive a file named fwdrequestingducdata.zip 2. Clone this repository by running the following command: git clone https://github.com/liatschiff/OpenAsp.git 3. Optionally create a conda or virtualenv environment: bash conda create -n openasp 'python>3.10,<3.11' conda activate openasp 4. Install python requirements, currently requires python3.8-3.10 (later python versions have issues with spacy) bash pip install -r requirements.txt 5. copy fwdrequestingducdata.zip into the OpenAsp repo directory 6. run the prepare script command: bash python prepare_openasp_dataset.py --nist-duc2001-user '<2001-user>' --nist-duc2001-password '<2001-pwd>' --nist-duc2006-user '<2006-user>' --nist-duc2006-password '<2006-pwd>' 7. load the dataset using huggingface datasets python from glob import glob import os import gzip import shutil from datasets import load_dataset openasp_files = os.path.join('openasp-v1', '*.jsonl.gz') data_files = { os.path.basename(fname).split('.')[0]: fname for fname in glob(openasp_files) } for ftype, fname in data_files.copy().items(): with gzip.open(fname, 'rb') as gz_file: with open(fname[:-3], 'wb') as output_file: shutil.copyfileobj(gz_file, output_file) data_files[ftype] = fname[:-3] # load OpenAsp as huggingface's dataset openasp = load_dataset('json', data_files=data_files) # print first sample from every split for split in ['train', 'valid', 'test']: sample = openasp[split][0] # print title, aspect_label, summary and documents for the sample title = sample['title'] aspect_label = sample['aspect_label'] summary = '\n'.join(sample['summary_text']) input_docs_text = ['\n'.join(d['text']) for d in sample['documents']] print('* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *') print(f'Sample from {split}\nSplit title={title}\nAspect label={aspect_label}') print(f'\naspect-based summary:\n {summary}') print('\ninput documents:\n') for i, doc_txt in enumerate(input_docs_text): print(f'---- doc #{i} ----') print(doc_txt[:256] + '...') print('* * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * * *\n\n\n') ### Troubleshooting 1. Dataset failed loading with load_dataset() - you may want to delete huggingface datasets cache folder 2. 401 Client Error: Unauthorized - you're DUC credentials are incorrect, please verify them (case sensitive, no extra spaces etc) 3. Dataset created but prints a warning about content verification - you may be using different version of NLTK or spacy model which affects the sentence tokenization process. You must use exact versions as pinned on requirements.txt. 4. IndexError: list index out of range - similar to (3), try to reinstall the requirements with exact package versions. ### Under The Hood The prepare_openasp_dataset.py script downloads DUC and Multi-News source files, uses sacrerouge package to prepare the datasets and uses the openasp_v1_dataset_metadata.json file to extract the relevant aspect summaries and compile the final OpenAsp dataset. ## License This repository, including the openasp_v1_dataset_metadata.json and prepare_openasp_dataset.py, are released under APACHE license. OpenAsp dataset summary and source document for each sample, which are generated by running the script, are licensed under the respective generic summarization dataset - Multi-News license and DUC license.

This is a dataset for multi-document summarization in Portuguese, what means that it has examples of multiple documents (input) related to human-written summaries (output). In particular, it has entries of multiple related texts from Brazilian websites about a subject, and the summary is the Portuguese Wikipedia lead section on the same subject (lead: the first section, i.e., summary, of any Wipedia article). Input texts were extracted from BrWac corpus, and the output from Brazilian Wikipedia dumps page. BrWac2Wiki contains 114.652 examples of (documents, wikipedia) pairs! So it is suitable for training and validating AI models for multi-document summarization in Portuguese. More information on the paper "PLSUM: Generating PT-BR Wikipedia by Summarizing Websites", by André Seidel Oliveira¹ and Anna Helena Reali Costa¹, that is going to be presented at ENIAC 2021. Our work is inspired by WikiSum, a similar dataset for the English language.

GameWikiSum is a domain-specific (video game) dataset for multi-document summarization, which is one hundred times larger than commonly used datasets, and in another domain than news. Input documents consist of long professional video game reviews as well as references of their gameplay sections in Wikipedia pages. Source: https://github.com/Diego999/GameWikiSum

In an active e-commerce environment, customers process a large number of reviews when deciding on whether to buy a product or not. Abstractive Multi-Review Summarization aims to assist users to efficiently consume the reviews that are the most relevant to them. We propose the first large-scale abstractive multi-review summarization dataset that leverages more than 17.9 billion raw reviews and uses novel aspect-alignment techniques based on aspect annotations. Furthermore, we demonstrate that one can generate higher-quality review summaries by using a novel aspect-alignment-based model. Results from both automatic and human evaluation show that the proposed dataset plus the innovative aspect-alignment model can generate high-quality and trustful review summaries. Paper: Large Scale Abstractive Multi-Review Summarization (LSARS) via Aspect Alignment