Quick Start

Figures and analyses reported below can be reproduced using the material in the exploratory data analysis folder.

EP data (1978-2019)

Number of EPO patents and patent families

Patent family

A patent family is a collection of patent applications covering the same or similar technical content. The applications in a family are related to each other through priority claims.

From now on, we work at the family level. Each family is assigned to the year of it first appearance in the EP full-text patent dataset. This date might differ from its priority year.

EP full-text data (1978-2019)

Number and share of EPO patent families with full-text data

Number Share
title abstract description claims amendment url
1978-2019 0.6 0.2 0.33 0.33 0.0 0.41

Overall, 1/3 of EP families have at least one patent with a full-text description.

Number and share of EPO patents and patent families with full-text data in english

Number Share
en_title en_abstract en_description en_claims en_amendment en_url
1978-2019 0.22 0.02 0.22 0.22 0.0 0.22

Overall, 1/5 of EP families have at least one patent with a full-text description in english.

Worldwide coverage of full-text patent data

Leverage patent families to enlarge full-text data coverage

Patents in the same (simple) family usually have the same text components. This means that any patent text can be propagated to all patents in the same family. Thanks to that, the EP (resp US) full text data can cover patents well beyond the the EP (resp US) office. We use this tip to look at the implicit worldwide coverage of the EP and US full-text patents.

In this section we focus on the availability of the full-text patent descriptions.

USPTO and EPO full-text patent description data coverage

USPTO full text data

The USPTO full text data are available from the USPTO bulk data. US full-text patent data can be easily parsed thanks to iamlemec/patents. They are also available in a well-structured format as part of the BigQuery patents-public-dataset.

USPTO EPO

Paradoxically, we observe that the USPTO full-text data implicitly covers a larger share of EP families than the EP bulk dataset.

The best of both worlds

We can also pool the families covered by the EPO and the USPTO full-text bulk datasets. This is the largest coverage of full-text patent data that can be obtained from publicly available datasets.