U.S. Congressional Speech Corpus Processed for NLP

A processed text corpus of U.S. parliamentary proceedings, likely containing congressional speeches or debates. The dataset was created by JonathanSum and uploaded to Hugging Face in February 2022. Its size is indicated by platform tags as being between 1 million and 10 million data points.

Use Cases

• Train a topic classification model on speech text to categorize debates by policy area.
• Perform named entity recognition to identify mentioned politicians, organizations, and locations within the proceedings.
• Analyze sentiment or polarity of parliamentary speech text over time.
• Use the text for training or fine-tuning large language models on formal political language.

Strengths

• Dataset size is categorized between 1 million and 10 million data points, suggesting a substantial collection.
• Data is processed and available in Parquet format, optimized for efficient analysis with libraries like Polars and Pandas.

Limitations

• Specific column names, row counts, and the exact source of the parliamentary text are unknown.
• Data is from 2022 or earlier, potentially missing recent congressional sessions and debates.

Provenance

Source

Likely derived from official U.S. congressional records (e.g., Congressional Record), but not explicitly stated.

Collection Method

Processed from source parliamentary text; specific cleaning and preprocessing steps are unknown.

Freshness

Last updated on 2022-02-17; no update frequency specified.

Geography

United States (inferred from 'Regionus' platform tag).