Datasets:

arbitrdatasets
/

arbitr-uspto-ev-sample

This repository contains a lightweight sample of the full arbitr-uspto-ev dataset so you can inspect the schema, directory layout, and licensing terms on Hugging Face before downloading the complete archive.

What's Inside

data/patents/patents_2025-10.parquet — 200-row slice from the Oct 2025 granted-patent export (PTGRDT).
data/applications/applications_2025-10.parquet — 200-row slice from the Oct 2025 application export (APPDT).
data/texts/texts_sample.jsonl.gz — same JSONL claims sample provided in the full dataset.
data/provenance/ — provenance summary + download log for this sample.
metadata/ — schemas, CPC whitelist, and machine-readable manifest matching the on-disk structure.
LICENSE_Arbitr_Commercial.txt — Arbitr Inc Commercial Dataset License v1.1 (required for redistribution).
CHANGELOG.md — release history for the sample package.

Purpose

Give Hugging Face users a quick preview of the normalized schema and CPC filtering.
Provide a minimal artifact (<1 MB) for automated tests.
Document licensing and provenance in the same layout as the production dataset.

Using the Sample

import pandas as pd

patents = pd.read_parquet("data/patents/patents_2025-10.parquet")
applications = pd.read_parquet("data/applications/applications_2025-10.parquet")

The schema is identical to the full release, so any pipelines built against this sample will work against the larger dataset (subject to licensing).

Getting the Full Dataset

Contact Arbitr Inc ([email protected]) or reference the main arbitr-uspto-ev package for production-sized exports covering 2015–2025 with CPC/keyword filters tuned for Electric Vehicle R&D.

License

See LICENSE_Arbitr_Commercial.txt. Public use must include:

© Arbitr Inc, 2025. Derived from public-domain sources (e.g., USPTO). No affiliation or endorsement by USPTO.