Alternative data datasets
Browse by provider, category, marketplace, access, delivery and license
A large granted-patent text collection from the USPTO, tens of millions of records in Parquet (Apache 2.0). Broad coverage of claims and descriptions for IP-intensity and technology-trend measurement.
About 1.5 million US patent claims split into training and test partitions in CSV, organized for claim-level classification and summarization work (Apache 2.0). A large, ready-split corpus for IP-claims modeling.
The patents-publications-dataset is a large-scale tabular collection of patent publication records distributed in Parquet format. Hosted by publisher labofsahil on the Hugging Face Hub under an unstated license, it is sized in the 100M–1B row range.
US patent full text assembled by the Allen Institute from the USPTO, millions of records under ODC-BY: claims, abstracts, descriptions and metadata, in Parquet. A broad base for measuring IP intensity, technology landscapes and the direction of R&D spending.
Google Patents Chemistry Terminology Benchmark This repository contains the completed google_patents_chemistry_250_luna_scratch benchmark: 4,500 segment pairs across 18 ordered translation directions, with 250 examples per direction. It is an evaluation resource, not model translation output. Generation completed on 12 October 2026. Files and Usage google-patents-18-directions-4500-pairs-manifest.
Mechanical Utility Patents — 100 HTML Records A curated collection of 100 distinct granted U.S. utility patents from 100 patent families, saved as HTML for mechanical-engineering document analysis. Published by cmuchancel. Source disclosure: the patent text, bibliographic metadata, and drawings were obtained from Google Patents. Original PDFs and drawing assets are hosted on patentimages.storage.g
Monitor the intellectual property filings that influence pharmaceutical innovation and competition.
A collection of 49,023 patent filings focused on distributed ledger technology is available, comprising roughly 1.296 billion tokens. The corpus is a component of the broader DLT-Corpus initiative and is intended to facilitate natural language processing research, innovation analysis, and patent examination within the distributed ledger field.
Raw French patent publications from 2020 through 2026, pulled from original A1 XML records by an independent API/FTP extraction and delivered one-document-per-row in streaming-ready parquet.
Google patents 2024–2026: an index of 4,651 patent families An AI model rated 339 Google and DeepMind inventions from 2024–2026 as directly about search. Only 93 of them have a write-up in any of the three sources that track Google patents for SEO. Roughly three in four are not covered there. This dataset lists all 4,651 patent families with a publication between January 1, 2024 and September 17,
google-patents-data-preview is a community-uploaded preview of Google Patents bibliographic records, distributed in Parquet format through the Hugging Face Hub. The single training split contains roughly 340,000 rows covering patent identifiers, classifications, and localized text fields.
A read-only archival repository serving as the official worldwide record for 'River of Cognition' technology patents and compliance authorizations, used for public filing and prior-art preservation of complete invention patent documents.
Original TIFF drawings and grant full-text XML for 165,917 U.S. design patents omitted from the AI4Patents/IMPACT collection, filling gaps including 161,093 grants from 2023 to 2026 and 4,824 earlier patents missing from IMPACT.
The us-patents dataset is a collection of ~ 8M US patent grants and applications from 1976-2025, cleaned, filtered, and formatted for pre-training of language models. Document Format corpus_id: Unique integer key with no semantic value. filing_date: The filing date of the grant or application. In case of duplicates, earliest filing date from the duplicate cluster. patent_type: The type of patent.
Disambiguated US patent data, inventors, assignees, locations, claims and citation networks, from PatentsView and Patent Examination Research, reaching back to 1976 and including pre-grant publications, all queryable for free.
Derived from USPTO and PatentsView bulk releases, this dataset catalogs 9.1 million U.S. patents from 1976 onward along with complete citation links, assignees, inventors, and CPC codes, with the full-graph edition exceeding 255 million rows.
ALEA Patent Claim Graph Reading questions about the dependency structure of real U.S. patent claim sets: which claim a dependent claim refers back to, the chain of claims it rests on, its independent claim, which and how many claims depend on a given claim, and yes / no questions about dependency. Every answer is computed deterministically from the claim text shown in the row. This dataset is part
Chinese patents paired with their abstractive summaries in Mandarin, a few thousand records in JSON (Apache 2.0). A window into the pace and direction of patenting inside the Chinese technology base.
Research and intelligence covering pharmaceutical products, clinical trials, patent filings and regulatory progress.
DAPFAM patent documents, a curated sample of US patents in Parquet with text for retrieval and classification tasks (CC BY-NC-SA 4.0). Scopus of claim text for building patent-similarity and technology-mapping models.
Intellectual-property datasets that capture patent claims, ownership chains and dispute records to support IP-driven strategies.
French patents published in 2024, split into chunks at the `<p>` XML tag level and saved in Parquet for streaming and distributed processing. The records come from public French patent filings (A1, 2024) and were independently extracted, structured, and segmented using access to PI API and FTP sources.
Patent analytics offering portfolio insights on innovation trends, assignee behaviour and citation patterns (IPQwery).
The Future of Patent Corpus Management One Document, Fully Decoded - Without Anyone Reading It The signal you cannot get from a search interface The highest-value competitor signal in a patent portfolio is not the invention. It is the amount of money and urgency a company committed to it. That signal is never in an abstract, and no search interface surfaces it. It comes out of this corpus for ever
Machine-readable extraction of financial and operating figures drawn from publicly filed corporate documents.
Patent All Claims EP granted claim sets (indepdendent and dependent claims). One claim per line. Claims from 20210915-20250806 Splits Train: 96% Validation: 2% Test: 2% Usage from datasets import load_dataset ds = load_dataset("mhurhangee/ep-patent-all-claims")
A collection featuring the primary independent claim of United States utility patents issued between 2005 and 2025 is provided. Because the opening claim usually establishes the widest legal boundaries of an invention, it carries particular significance, and each entry includes the patent identifier alongside its claim text.
Over 1.3 million granted US patents from the BIGPATENT benchmark, each pairing its claims with an abstractive summary and full description text in English (CC BY 4.0, Parquet). Use it to track the technology areas where applicants are filing and to study how claim language has evolved over time.
Patent-linked intelligence that connects intellectual property activity to research strategy and competitive standing.
PatentPulse PatentPulse is a provenance-preserving corpus of USPTO grants and published patent applications extracted from official weekly XML bulk releases. This immutable Parquet snapshot normalizes the project's historical append-only JSONL into one schema. Historical partial snapshot. This release was captured on 2026-08-28 while the upstream backfill was still in progress. It is a stable, cit
Qwen3-4B Patent Embeddings (专利特征向量数据集) 📌 简介 (Introduction) 本数据集包含了海量专利文本的特征向量（Embeddings）。所有特征向量均使用强大的 Qwen3-embedding-4B 模型提取。 数据领域: 专利 (Patents) 特征提取模型: Qwen3-embedding-4B 数据格式: .mmap (Memory-mapped file, numpy-compatible) + .tsv 数据精度: float16 📂 数据结构 (Data Structure) 由于数据量庞大，在生成时使用了 4 张 GPU 进行分布式计算，因此数据被划分为 4 个分片（Shards）：shard_0 到 shard_3。 每个 Shard 目录下包含以下核心文件： embeddings.float16.mmap: 核心的特征向量矩阵
A collection of Parquet files providing one record per Chinese invention patent family granted between 2003 and 2019, capturing a cumulative quality assessment pipeline that includes metrics such as semantic knowledge recombination, strict historical semantic novelty, and IPC-based robustness measures across global patent publications.