Cite the Record
Menu
science analytics database

For academic researchers

Full provenance, every claim citable.

Live search and source-linked record pages are available now. Structured data endpoints, a direct AI-tool connection and weekly source-family files are planned but not yet deployed. Supported records expose available provenance metadata.

Available and planned research surfaces

Pick the one that fits your workflow. They all read from the same canonical source.

Available search and structured records, compared with planned direct-tool and source-family file access.
Surface Best for Format Latency Reproducibility
Direct connection (planned) Interactive research sessions in Claude / Cursor / your agent. HTTP+SSE / JSON-RPC Planned; not deployed Pin parser version + timestamp; record returned evidence_sha256.
Live search + record pages One-off discovery, reading and source verification. Server-rendered HTML Available now Retain stable identifiers, official links and available provenance from the record page.
Source-family files Batch processing, NLP corpora, fine-tuning, longitudinal study. Machine-readable file Planned weekly cadence Versioned publication is planned.

What you'll use most

  • Planned weekly source-family files — versioned publication for reproducible research is being onboarded. Browse source families.
  • JSON Schema reference — planned shard kinds have versioned field tables; canonical schema files and verified examples are pending. Browse schemas.
  • Planned direct AI-tool connection — read-only search, retrieval and citation tools are coming later and are not yet deployed. Read the walkthrough.
  • Available provenance on live record pages — stable identifiers, official source URLs and retained-source metadata where present.
  • Citation controls — BibTeX, APA 7, Hansard-style and defensive-footnote formats on supported record pages. Format reference.
  • Coverage boundary documented — the platform is honest about which Parliaments are in scope, which source families are covered, and which are roadmap. Current coverage.

A typical workflow

  1. Define your corpus with live search and check the coverage boundary. Source-family pages document planned publication but do not provide downloads yet.
  2. Collect records through live search and record pages, checking the documented coverage boundary.
  3. Run your analysis in your tooling of choice using the record text and metadata actually present.
  4. Verify edge cases against the official-source link and retained evidence metadata.
  5. Cite your sources with the format generator on a live record page.
  6. Use versioned source-family files later when that publication surface is onboarded.

Reproducibility and the provenance contract

Live records carry the identifiers needed to verify them against retained source bytes. Future data files and direct-tool responses will preserve the same provenance contract. If an upstream source changes a record, the platform records the change event without rewriting prior captured versions.

For methodology citations, the About page documents the parser versions, the storage philosophy, the neutrality contract, and the maintainer-attestation chain.

Quick-look an evidence pack

A topic evidence pack is designed to provide a structured view of recent records so you can check corpus scope, parser output and the citation chain when topic onboarding is complete.

Limits and honesty

The platform does NOT publish derived inference fields — no sentiment scores, no influence rankings, no actor-network classifications. If your research requires such derivations, you build them; the platform refuses to ship them so consumers can't accidentally inherit our framings. See the neutrality policy.

Coverage is bounded — see coverage. Historical Parliaments and devolved legislatures are not yet ingested. If your study depends on them, contact feedback@bower.im to flag the dependency.

Start here