Technological weight for CPC (CPC/IPC concordance)

[VanessaBehrens!]

Dear EPO PATSTAT support,

I would like to have a weight for the share of cpc technological fields (similar to the "weight" variable from TLS230_appln_techn_field but for CPC). I tried to use the concordance table (https://publication.epo.org/raw-data/pr ... oductId=71) but the realised that that does not count the claims.

I then thought of counting the number of times a given cpc_class_symbol occurs per appln_id and divide by the sum of cpc_class_symbols for each appln_id. However, I assume that the cpc_class_symbols listed per appln_id are unique and therefore counting them would not be informative as to how many times they occured within that appln_id.

Do you have any ideas on how I could proceed?

Best wishes,
Vanessa

[EPO / PATSTAT Support]

Dear Vanessa,
CPC classifications are assigned at family level and you will not find duplicates. (CPC is also not -yet-published on the published patent document.)
IPC is a bit different, they are published and can change between the publication instances. We group them and de-duplicate them at application level. Additionally, when patents (or rather said groups of patents) are re-classified, the whole simple docdb family will receive these new IPC codes.
This will lead to IPC codes being different between what is printed on paper, and what is available in PATSTAT. But it has the great advantage of grouping all patents belonging to the same technology into the same classes, independent from when the patent was published, searched or granted.
I think the only way of assigning a technological weight based on CPC (or IPC) would be similar to what was done to create the industry(NACE) and technology mapping. And this in its turn would lead to looking at IPC/CPC at 4 digit (or lower) level. Basically, it would mean re-doing the current technology mapping to your needs, develop your own taxonomy. (Similar to the EPO Y-tags, or the OECDs' J-TAG for ICT mapping

OECD ICT-A new taxonomy.pdf

(919.15 KiB) Downloaded 192 times

)
It might be feasible to do this for a specific area (such as Block Chain or AI) - but for sure a lot of work and thinking-.

Geert BOEDT