Loading...
Loading...
News corpora, social media analysis, movie/music metadata, sports data, cultural datasets, misinformation
11,250 datasets
30,000 messages categorized into 36 disaster response classes, sourced from major global events like the 2010 Haiti earthquake and 2012 Super-storm Sandy. The collection includes direct communications and news articles spanning multiple years and hundreds of distinct disaster scenarios.
Two news sources, Hamshahri and RadioFarda, contribute Farsi (Persian) articles extracted from RSS feeds. The text data captures headlines and summaries, representing both domestic Iranian and international media perspectives.
Two categories of English news article URLs, fake and satire, are paired with rebuttal links for the fake entries. The dataset utilizes article_number and url_of_article fields to organize the source material for misinformation analysis.
3,126 Swahili news articles categorized into 6 distinct classes including National, International, and Business. The collection represents media from Tanzania, where the language serves as the primary medium of instruction for 100-150 million speakers.
Urdu language news articles organized into 5 intent categories for veracity classification. It provides text data in the `news` column alongside `label` and `category` fields to support misinformation detection in a low-resource language.
A machine-translated version of the OpenAssistant/oasst1 dataset into the Shan language, created using the facebook/nllb-200-3.3B model. The dataset's author is NorHsangPha, and it was last updated on July 11, 2024. The description notes that the data quality has not been verified by human reviewers.
833 parallel language pairs of news titles covering 64 languages and 17 scripts, created by aiana94 and last updated in June 2024. PolyNewsParallel is a multilingual dataset designed for cross-lingual tasks. The dataset page indicates it contains news titles, but specific column details and data volume are not provided.
Designed for named entity recognition tasks, containing text in multiple languages including Dutch, French, and German. The annotations were generated by experts, and the data is stored in Parquet format. Specific details on row count, column structure, and data instances are unavailable.
2,909,551 news articles from the SogouCA and SogouCS corpora categorized into 5 classes. The dataset features 450,000 training and 60,000 testing samples where original Chinese characters are converted to Pinyin.
A dataset titled 'Indian Actors Film' was published on the HuggingFace platform by SuperNived on August 1, 2024. The dataset likely contains information related to actors in the Indian film industry, but its specific content, size, and structure are unknown from the provided metadata.
CSMeD is a collection of datasets for automated citation screening in systematic literature reviews, introduced by Wojciech Kusa at NeurIPS 2023. It provides labeled bibliographic data derived from Cochrane reviews to support the development of NLP models for medical research synthesis. The collection focuses on bridging the gap between manual screening processes and machine learning automation.
5,300 annotated news documents and 38,000 paragraphs containing sentiment labels for 3,200 unique target entities. The dataset captures author-specific sentiment towards a central subject across both full-text and individual paragraph contexts.
Polish news articles and their corresponding summaries organized into positive and negative pairs. It utilizes summaries of the same article as positive matches and samples the most semantically similar summaries from different articles to serve as hard negatives for contrastive learning.
2.7 terabytes of data representing a dump of Reddit content spanning nearly two decades. The torrent was compiled by the user 'nothingiisreal' and shared via Academic Torrents. Its large scale and long time range make it a notable resource for language model research.
Encompassing 36,591 movie reviews from the Kinopoisk platform, spanning from July 2004 to November 2012. It includes reviews from the TOP250 and BOTTOM100 rank lists, with sentiment scores labeled as Good, Bad, or Neutral.
A text dataset from the Hugging Face platform, likely containing language data for detecting signs of suicide and depression. The dataset was published by author 'joshyii' and was last updated on July 30, 2024. Its specific size, geographic origin, and detailed content are not described in the available metadata.
A Web Map Service (WMS) provides the urban development plan ‘Sporthalle Ditzingen’ for the city of Ditzingen. The data originates from the XPlanung 5.0 standard and is published by the Bundesamt für Kartographie und Geodäsie. It was last updated on June 13, 2024.
10 categories, including Politics, Economy, and Technology, classify Uzbek news texts. The dataset was created by risqaliyevds and published on Hugging Face in June 2024. It is structured in JSON format for zero-shot classification tasks.
A corpus of approximately 2 billion tokens of Traditional Chinese text, processed for language model performance. The dataset, created by liswei and last updated on June 2, 2024, contains high-quality news and articles focused on Taiwanese culture. It has been de-duplicated and cleaned using rule-based and learning-based filters.
World trade, 1800-1938: a new data-set presents import and export series for 149 polities from 1800 to 1938. The dataset, created by Giovanni Federico and Antonio Tena Junguito, includes values at current and constant (1913) prices and borders. It is related to a working paper available via the provided handle.