User:Econterms/WikiProject Patents

Lightning talk presentation to WikiConference North America 2018

• Wikidata can record basic information (not detailed information) about tens of millions of patents, someday. Right now there are only a few hundred.
• Project's goal: set standards for patent data on Wikidata, and make it easy
• The WikiProject Patents page: Wikidata's WikiProject Patents

• Focus: patents from before 1923, because
  • They're beyond copyright
  • Their claims (almost?) never apply any more
  • Patents were shorter and simpler back then
  • There are not as many: Fewer than 100K annually worldwide before 1910. The numbers grew exponentially. Now, 3 million a year, on the order of 9,000 a day.
  • This is relevant to my off-wiki research, tracking aero technology back then
  • There exists a lot of specialized software to manage the most recent patents, which are relevant to industry today
• I've begun a conversation with WIPO (World Intellectual Property Organization, the UN unit that manages the more recent treaty's relationships)

Patent data elements

• Instance of (P31) A patent item should be an instance of either patent (Q253623) or U.S. Patent (Q43305660), perhaps both. That property is the one to query (search) that is unique to patents.
• Page title -- one standard form: Patent US-1906-827017, Patent CA-1914-153820 -- different titles are fine too
• Country where filed: Here are three options; freely use any or all. They express slightly different things
  • Use issued by (P2378) and identify the office to which the patent was filed -- e.g. US Patent and Trademark Office, Japan Patent office (JPO)
  • Or, "applies to jurisdiction" (P1001) and then the Q-id of the government; or, country (P17) and then the Q-id of the national government/country. The country may not still exist.
• Filing date: Formal date of submission of the patent application, and generally speaking the date on which the patent goes into force legally once it's approved
• Grant date: Certification by a government that the patent is accepted, and applies in the jurisdiction. (Might be more complicated with later international treaties.)
• Applicant(s) -- there's always at least one ; can include company or university or government lab
• Inventors: Zero or more; Might like to mark their order for some we have "author name strings", for others Q-ids (same for scientific publications)
• Title: Applicants give a title in the language of t
• Patent number -- inherited from years ago, e.g. US821393 -- works for those on google patents, and automatically links to that source
  • PROBLEM: too strict a format ; what to do for the ones that don't fit the format?

• link to Wikisource if patent document is there
• Link to Q-id or string of Parent patent or child patent ?
• Assignee? Important in industrialization
• Pointer to URL with more information, possibly the full text and diagrams -- There is not yet a site and covers the 19th century completely. Wikidata could be the best site for this, someday.

Possible good outcome from getting patents onto Wikdiata

• We could add patent offices to the Authority Control line, maybe -- like USPTO, or WIPO, and if user clicks can get to a list of patents on Wikidata

• Link together patents transcribed on Wikisource
• Chart patent counts by inventor, country, tech topic; Time lines
• Other insights?

Next steps

• There are a few hundred patents on Wikidata. I will upload more, probably QuickStatements (thanks to Jarekt's help), still just a few
• Here's the QuickStatements: https://tools.wmflabs.org/quickstatements/#/batch

 CREATE								
 LAST	Len	"Patent US-1906-827017"						
 LAST	P31	Q253623						
 LAST	P1476	en:"Wing of flying machines"	S1246	"US827017"	S813	+2018-10-19T00:00:00Z/11	S248	Q3235742

• Any input? How should this be done? What would be useful to you?