CNN DailyMail News Articles With Coreference Annotations

A text dataset based on the CNN/DailyMail news summarization corpus, enhanced with coreference annotations. The dataset was created by user 'cestwc' and uploaded to Hugging Face in May 2022. It is designed for training and evaluating models on the task of resolving pronoun references within news narratives.

Use Cases

• Train coreference resolution models to link pronouns like 'he', 'she', or 'it' to their correct antecedent named entities within article sentences.
• Evaluate the impact of resolved coreference chains on the quality of generated abstractive summaries of news articles.
• Analyze patterns of entity mentions and pronoun usage across different news domains and article lengths.
• Fine-tune language models on a structured discourse task using annotated news text data.

Strengths

• Built upon the established CNN/DailyMail corpus, a benchmark for summarization containing over 300,000 news articles.
• Provides explicit annotations for a fundamental and challenging NLP task beyond plain text.
• Hosted on a major platform (Hugging Face) with supporting libraries like datasets and dask for accessibility.

Limitations

• The core annotation quality and methodology are unknown without examining the source code or paper.
• Dataset size, specific annotation statistics, and license details are not provided in the metadata.
• Last update was in 2022, which may not reflect recent advancements in coreference annotation schemas.

Provenance

Source

Derived from the CNN/DailyMail news summarization dataset.

Collection Method

Annotations added to the original articles; specific annotation tool or method is unknown.

Time Range

Temporal coverage of original CNN/DailyMail articles.

Freshness

Last updated 2022-05 - 22; update frequency unknown.

Geography

News content primarily from US sources (CNN, Daily Mail).