yeeder/congressional-record-parquet

Congressional Record 43rd–114th Congresses — Parquet Edition

This repository contains a Parquet-formatted derivative of the Stanford Congressional Record for the 43rd–114th Congresses: Parsed Speeches and Phrase Counts dataset.

The files were prepared as a compact, query-friendly research corpus for historical text search with tools such as DuckDB and the Congressional Record Explorer.

Coverage

• Congresses: 43rd–114th
• One Parquet file per Congress
• Files: data/congress_043.parquet through data/congress_114.parquet
• Language: English

The exact row counts, year ranges, and file sizes for each Congress are recorded in manifest.json.

Source dataset

This repository is derived from:

Matthew Gentzkow, Jesse M. Shapiro, and Matt Taddy.
Congressional Record for the 43rd-114th Congresses: Parsed Speeches and Phrase Counts.
Palo Alto, CA: Stanford Libraries [distributor], 2018-01-16.

Original project page:
https://data.stanford.edu/congress_text

License and attribution

The source dataset is distributed under the Open Data Commons Attribution License (ODC-BY 1.0).

This repository is a derived representation of that dataset and retains the source attribution. Users should review the original dataset documentation and license when redistributing or reusing these files.

Processing

The Stanford source archives were transformed into Congress-level Parquet files.

Each speech record includes the speech text and available metadata such as:

• speech ID
• Congress
• date and year
• chamber
• speaker ID and speaker name
• state
• gender
• party
• district
• nonvoting status
• character count
• word count
• source identifier

The transformation joins Stanford speech, description, and speaker-map records by speech_id.

No topical, religious, semantic, or keyword filtering was applied to the text in this full Parquet corpus.

File organization

data/
├── congress_043.parquet
├── congress_044.parquet
├── ...
├── congress_113.parquet
└── congress_114.parquet

Research use

The Parquet representation is intended to support reproducible historical research, including:

• literal phrase search
• keyword and term-combination search
• filtering by year, Congress, chamber, party, state, and speaker
• chronological comparison
• later lexical or semantic retrieval experiments

Parquet is the processed research format. The original Stanford archives remain the authoritative upstream source.

Reproducibility

For reproducible work, record both:

1. the citation for the original Stanford dataset; and
2. the Hugging Face repository revision used in the analysis.

For example, the first complete Parquet upload was committed at revision:

2df2309ae4e86f7c112e91f94912ecddbeb0188a

Later revisions may update documentation, manifests, or processed files.

Citation

Please cite the original Stanford dataset:

@dataset{gentzkow2018congressionalrecord,
  author    = {Matthew Gentzkow and Jesse M. Shapiro and Matt Taddy},
  title     = {Congressional Record for the 43rd-114th Congresses:
               Parsed Speeches and Phrase Counts},
  publisher = {Stanford Libraries},
  address   = {Palo Alto, CA},
  year      = {2018},
  url       = {https://data.stanford.edu/congress_text}
}

When using this Parquet representation, also record the repository and revision:

Hugging Face dataset:
yeeder/congressional-record-parquet

https://huggingface.co/datasets/yeeder/congressional-record-parquet

Revision used:
<commit hash>

A citation for the derived representation can be written as:

@misc{congressional_record_parquet,
  title        = {Congressional Record for the 43rd--114th Congresses:
                  Parquet Edition},
  year         = {2026},
  howpublished = {Hugging Face Dataset Repository},
  url          = {https://huggingface.co/datasets/yeeder/congressional-record-parquet},
  note         = {Derived from Gentzkow, Shapiro, and Taddy (2018).
                  Record the repository revision used for reproducibility.}
}

Notes

This is a processed derivative intended for research convenience. Users who require the original archival organization or Stanford-provided derived phrase-count files should consult the upstream Stanford dataset.