Understanding data

[yoheikikuta]

TARGET DATA

Bulk Data Storage System: https://bulkdata.uspto.gov/#pats From this system we use the following datasets for our analysis.

Patent Application Full Text Data (No Images) (MAR 15, 2001 - PRESENT)

This dataset is composed of xml files (these are zip compressed) of each week's publication. Each xml file is as follows:

$ head -n 20 ipa170105.xml
<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE us-patent-application SYSTEM "us-patent-application-v44-2014-04-03.dtd" [ ]>
<us-patent-application lang="EN" dtd-version="v4.4 2014-04-03" file="US20170000001A1-20170105.XML" status="PRODUCTION" id="us-patent-application" country="US" date-produced="20161220" date-publ="20170105">
<us-bibliographic-data-application lang="EN" country="US">
<publication-reference>
<document-id>
<country>US</country>
<doc-number>20170000001</doc-number>
<kind>A1</kind>
<date>20170105</date>
</document-id>
</publication-reference>
<application-reference appl-type="utility">
<document-id>
<country>US</country>
<doc-number>14789882</doc-number>
<date>20150701</date>
</document-id>
</application-reference>
<us-application-series-code>14</us-application-series-code>

Important tags:

• publication-reference kind:
  • A1: Patent Application Publication
  • A2: Patent Application Publication (Republication)
  • A9: Patent Application Publication (Corrected Publication)
• publication-reference date: publication date (typically 1.5 years from application)
• application-reference doc-number: app_id
• application-reference date: application date

Patent Grant Full Text Data (No Images) (JAN 1976 - PRESENT)

This dataset is composed of xml files (these are zip compressed) of each week's granted patent. Each xml file is as follows:

$ head -n 20 ipg120103.xml
<?xml version="1.0" encoding="UTF-8"?>
<!DOCTYPE us-patent-grant SYSTEM "us-patent-grant-v42-2006-08-23.dtd" [ ]>
<us-patent-grant lang="EN" dtd-version="v4.2 2006-08-23" file="USD0651376-20120103.XML" status="PRODUCTION" id="us-patent-grant" country="US" date-produced="20111219" date-publ="20120103">
<us-bibliographic-data-grant>
<publication-reference>
<document-id>
<country>US</country>
<doc-number>D0651376</doc-number>
<kind>S1</kind>
<date>20120103</date>
</document-id>
</publication-reference>
<application-reference appl-type="design">
<document-id>
<country>US</country>
<doc-number>29390372</doc-number>
<date>20110423</date>
</document-id>
</application-reference>
<us-application-series-code>29</us-application-series-code>

Important tags:

• application-reference doc-number: granted patent number
  • Start from 'P': Plants (品種登録)
  • Start from 'D': Design (意匠)
  • Start from 'R': re-issue (再発行)
  • Start from '0': Patent (特許) ※ Not Yet confirmed
• publication-reference kind (see Kind Codes list for detail):
  • For doc-number starting from 'P': P1, P2, P3, P4, P9
  • For doc-number starting from 'D': S
  • For doc-number starting from 'R': R
  • For doc-number starting from 'B':
  • B1: Patent | No previously published pre-grant publication
  • B2: Patent | Having a previously published pre-grant publication and available March 2001
• publication-reference date: publication date of a granted patent

Patent Application Office Actions Research Dataset (Stata (.dta) and MS Excel (.csv)) (2008 - JUN 2017)

URL: https://bulkdata.uspto.gov/data/patent/office/actions/bigdata/2017/ See this pdf for detail.

• citations.csv.zip
  • app_id: Application number
  • citation_pat_pgpub_id: Patent or pre-grant publication prior art cited
  • parsed: Patent or pre-grant publication number parsed from citation_pat_pgpub_id
• office_actions.csv.zip
  • app_id: Application number
  • ifw_number: Image File Wrapper (IFW) number of the Office action
  • rejection_xxx
  • 101: Subject Matter Eligibility, Statutory Double Patenting, Utility, etc
  • 102: Lack of Novelty
  • 103: Obviousness
  • 112: Written Description, Indefinite Claims, etc

[yoheikikuta]

How to use these datasets:

• (Pick up {application data, grant data, citations and office_action} in specific years)
  • All data is so huge...
• Create the combination of app_id and granted patent number that have the relationship of cite-cited (granted patents can be cited in applications)
  • Use citations.csv (app_id and parsed columns)
  • (In principle we can create the combination using only xml files)
• Extract patent text data of the created app_id and granted patent number
  • This step requires fast text processing (maybe Python is not a good way...)
• Analyze as we like