Datasets:
Add German EPO EP full text for text analytics
German patent descriptions, claims and abstracts from EPO’s specifically licensed text-analytics product.
Source: https://link.epo.org/web/ep_full_text_data_for_text_analytics-user_guide_v1.4_en.pdf
Licence: CC BY 4.0 (this product only).
Approx. tokens: ~30–70B across all languages; german share unmeasured.
The estimate is for the raw multilingual snapshot; measure the selected language during import. The former download bucket currently returns 404.
Measured the download before building this, and the product as described is no longer distributed:
- the bucket the user guide names,
ep-fulltext-for-text-analytics, answersNoSuchBucketfor the project, the object-list and the XML API, so theEPnn00000.txtfiles are gone; - the EPO product page the guide links to no longer exists - the bulk-data section now lists "EP full-text data" only;
- what EPO does publish is EP full-text data: XML ST36 and PDF/A, about 3.53 TB zipped for the backfile plus a 4 GB weekly frontfile, under the EPO database licence terms rather than the CC BY 4.0 of the text-analytics product.
3.53 TB does not fit this machine (790 GB free), and the remaining product is a different licence, so I am not starting it. The German share never became measurable because there was nothing left to measure. If you know a mirror of the text-analytics files, say so and I will take it up again; otherwise I will close this one as not deliverable.