Text / AI & Machine Learning
Keeper Vectorised Patent Data | Global Patent Embeddings for AI, LLM & Machine Learning Applications
Advertised by providerLicensing rights not independently verifiedPrice on request
Overview
Keeper Vectorised Patent Data | Global Patent Embeddings for AI, LLM & Machine Learning Applications is a text dataset. Typical applications include machine translation, LLM pre-training and agents & tool use. Relevant sectors: AI & Machine Learning and Education. Pricing is not published and is provided on request.
Primary applications
Data as of Q2 2026. The information on this website reflects the latest available data as of June 2026 and may not reflect subsequent changes. 152M+ full-text documents. 83 authorities. One feed. The world's most complete and most global patent collection, sourced directly from each publishing authority and delivered in your format, weekly — now vectorised for fast, accurate retrieval. This product pairs that collection with an efficient two-stage retrieval design so you can search and match across the entire corpus at scale. Why complete coverage matters A blind spot in patent data is a legal and commercial risk. Prior art your search universe doesn't contain might as well not exist until it surfaces in litigation, an opposition, or a competitor's hands. Coverage limited to the ~50 authorities that patent offices conveniently provide creates a false sense of security: innovation doesn't stop at those borders, and neither do the filings that can invalidate yours. The design goal of this collection is every authority we can reach, as deep into history as the records go, with the full document — not a stub. Coverage and depth - Complete coverage of the IP5 and PCT, plus 160+ further authorities most providers never reach (China 55.4M+, Japan 23.4M, United States 22M, Korea 7.8M, EPO 7.3M, PCT/WO 5.5M). - Full documents, not bibliographic stubs: original PDFs (159M), searchable claims (152M), descriptions (152M), images (130M), with English machine translation across the collect
Related datasets
More in Multimodal & Agent Training →TextAdvertised
Multilingual Middle East/North African News Events API: ML-Labelled, Arabic/Turkish/Farsi/English
Overview This API delivers structured, machine-labelled news events from the Middle East and North Africa as a live, queryable feed rather than a static download.
- Language
- Arabic, English +1
- Volume
- —
- Domain
- AI & Machine Learning +2
- Use case
- LLM pre-training
USD 499