Dataset
DAPFAM_patent
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.
Data preview
Dataset card
DAPFAM_patent is a text retrieval benchmark of patent documents, published on the Hugging Face Hub by datalyes, designed to evaluate cross-domain patent family retrieval models. It is distributed as a single training-style split with roughly 96,000 rows under a CC BY-NC-SA 4.0 license.
What this dataset contains
The dataset consists of patent records used to construct 18 retrieval tasks spanning in-domain, out-of-domain, and combined document/query views, with each task pairing titles and abstracts against titles, abstracts, claims, or full text. Text fields such as title_en, abstract_en, and claims_text describe each invention, while metadata fields such as earliest_claim_date, earliest_claim_year, earliest_claim_jusrisdiction, jurisdiction, ipcr_codes_str, and classifications_ipcr_list_first_three_chars_list support domain and jurisdiction filtering. The provided split contains 96,452 rows in parquet format, and an accompanying notes section indicates that a Snowflake Arctic embedding model was used as the reference encoder in the source paper.
Structure and fields
Each row is keyed by a relevant_id large_string and carries a timestamp earliest_claim_date together with string fields earliest_claim_year, earliest_claim_jusrisdiction, title_en, abstract_en, claims_text, description_en, ipcr_codes_str, and jurisdiction. Categorical arrays jurisdiction and classifications_ipcr_list_first_three_chars_list hold repeated string values for filtering by country and IPC technology class. The schema is text-only with no numeric price, flow, or transaction columns, and the dataset is exposed via the datasets, pandas, polars, and mlcroissant libraries.
How analysts use it
The intended use is benchmarking sentence-transformer style encoders on cross-domain patent retrieval through MTEB, rather than producing trading signals, so its direct applicability to investment workflows is limited. Analysts in patent-heavy sectors could repurpose title_en, abstract_en, claims_text, ipcr_codes_str, earliest_claim_date, and jurisdiction to monitor filing trends by technology class or country, but the source documentation frames the resource strictly as a research benchmark and provides no performance attribution or market outcomes.
Documentation & references 1
Related datasets
Browse moreParsed eligibility criteria from 13,229 ClinicalTrials.gov studies representing the candidate pool gathered by two first-stage retrievers during TREC Clinical Trials 2021–2023, formatted as typed entity-relation graphs for reranking models.
The clinical-trials-patient-graphs dataset is a small tabular collection of patient-level clinical records distributed in Parquet format by the 2001jdev publisher. It organizes entities and relations extracted from trial data into structured rows spanning 2021 through 2023.
This dataset is a parsed snapshot of ClinicalTrials.gov records, distributed in Parquet format on the Hugging Face Hub under an unspecified license. It contains roughly 52,000 trial entries drawn from public registry filings.
clinical-trials-trec-qrels is a tabular relevance-judgment file distributed by the 2001jdev user on the Hugging Face Hub. It maps clinical-trial topic identifiers to NCT registry IDs with graded relevance scores.
The clinical-trials-trec-topics dataset is a small tabular collection of clinical-trial topic records distributed in Parquet format on the Hugging Face Hub. It appears to be derived from TREC clinical-trial retrieval topics, with rows partitioned by year.
SEC EDGAR 10-K & 10-Q financial statements for 2,600+ NASDAQ companies.
Discussion
Questions or observations about this dataset? Join the conversation on X.
No comments yet. Start the conversation on X.