U.S. Congressional Speech Corpus from Parliamentary Processes

Parliamentary process transcripts from the U.S. Congress, compiled into a text corpus by M-Quan and hosted on HuggingFace. The dataset contains over 1 million to 10 million text entries, as indicated by its platform size category, and was last updated in March 2022.

Use Cases

• Train language models to identify procedural motions like 'point of order' or 'yield' within speech transcripts.
• Classify speech segments by speaker role (e.g., chairperson, representative) using metadata annotations.
• Analyze sentiment or topic evolution across congressional sessions using the chronological text data.
• Extract named entities such as bill numbers, agencies, or political figures mentioned in the proceedings.

Strengths

• Corpus size estimated between 1 million and 10 million text entries.
• Data is stored in Parquet format for efficient columnar access and processing.

Limitations

• Specific column names, row counts, and license information are unavailable.
• Data is from 2022 and may not include recent congressional sessions.

Provenance

Source

U.S. Congress parliamentary processes.

Collection Method

Text transcripts compiled into a corpus by M-Quan.

Time Range

null

Freshness

Last updated March 2022; update frequency unknown.

Geography

United States (Region:us).