Extreme Summarization Dataset for English Conversations

Featuring text data for the task of extreme summarization, as described in the EMNLP 2018 paper 'Don't Give Me the Details, Just the Summary!'. It is a monolingual English dataset with expert-generated annotations for text generation and summarization tasks. The dataset size is categorized as between 100K and 1M entries.

Use Cases

• Train a sequence-to-sequence model for extreme summarization using the expert-generated conversation and summary pairs.
• Fine-tune a transformer model on the monolingual English text data for abstractive summarization tasks.
• Analyze the relationship between conversation length and summary length in expert-generated annotations.
• Benchmark summarization models on a dataset specifically designed for extreme compression of text.

Strengths

• Dataset size is categorized as between 100,000 and 1,000,000 entries.
• Annotations are expert-generated, not crowdsourced.
• Data is monolingual English, providing a focused resource for that language.
• Associated with a peer-reviewed EMNLP 2018 research paper.

Limitations

• Specific row count, column names, and file formats are unknown from the input.
• The temporal and geographic coverage of the source conversations is not specified.
• The license is listed as 'cc By Nc Nd 4.0', which restricts commercial and derivative use.

Provenance

Source

Original dataset from the EMNLP 2018 paper authored by knkarthick.

Collection Method

Expert-generated annotations for summarization, as per the paper.

Freshness

Last updated on the platform was 2022-12-07.

Geography

Region is tagged as 'us', but specific spatial coverage is unknown.