Media Interview Transcripts with Summaries from NPR and CNN

463.6K interview transcripts paired with abstractive summaries. It was collected from NPR and CNN by author nbroad and last updated in October 2022.

Use Cases

• Train abstractive summarization models on the provided transcript-summary pairs.
• Analyze linguistic patterns and topic coverage in media interview transcripts.
• Fine-tune language models for generating summaries from long-form conversational text.
• Study the structure and content of interview data from major US news sources.

Strengths

• Contains 463.6K data points, providing a large-scale resource for text processing.
• Includes abstractive summaries for each transcript, offering direct supervision for summarization tasks.
• Sourced from established media organizations NPR and CNN, indicating a defined content domain.

Limitations

• The dataset is monolingual (English) and regionally focused on the US, limiting cross-lingual or global applications.
• The specific column structure and data schema are unknown, which may complicate initial exploration.
• The last update was in 2022, so the content may not reflect recent events or media trends.

Provenance

Source

NPR and CNN

Collection Method

Collected from interview transcripts and overview/topic descriptions.

Freshness

Last updated 2022-10-25.

Geography

United States (inferred from Regionus tag)