IPRoduct: A database of patent-product pairs OECD Blue Sky III Forum Ghent, Belgium, September 19–21, 2016 Gaétan de Rassenfosse Assistant Professor in Innovation and IP Policy Ecole polytechnique fédérale de Lausanne (EPFL), Switzerland @gderasse wledge financial support from the U.S. NSF (SMA - 1645264)
20
Embed
Rassenfosse - IProduct database of patent products pairs
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
IPRoduct:A database of patent-product pairs
OECD Blue Sky III ForumGhent, Belgium, September 19–21, 2016
Gaétan de RassenfosseAssistant Professor in Innovation and IP Policy
Ecole polytechnique fédérale de Lausanne (EPFL), Switzerland
@gderasse
I acknowledge financial support from the U.S. NSF (SMA - 1645264)
IPRoduct: A database of patent-product pairs 2
We are not able to trace innovations to the marketplace
Innovations are intangibles and, hence, difficult to study.
Scholars have focused on patents (and scientific publications and trademarks) as “output” of the innovation process.
But these are (i) abstract manifestations of science; (ii) intermediate innovation output; (iii) not tied to market outcome.
As a result, innovation studies have largely failed to study the “real impact” of innovation. Exceptions exist, such as studies on drug patents or case studies of particular technologies.
I want to observe innovations at the point at which they reach consumers
IPRoduct: A database of patent-product pairs
The holy grail of innovation research
I propose to build a table that links IPR data to product data.
Physical patent marking means that it is difficult to collect such information on a large scale.
But companies can now use virtual patent marking.
Allows them to collect damages for patent infringement with respect to infringing activity that occurs before the infringer is put on actual notice of the infringement.
5
The idea is simple: collect information available on line
IPRoduct: A database of patent-product pairs 6
But implementation is quite challenging
Several technical issues for collecting and structuring the data:– Format ranging from straight HTML files listing all products
and the associated patents to non-OCR PDF files– Information buried in the deep web (dynamic pages and
forms)– Product lines rather than actual products– etc.
But great technology is available. We are developing a software to do the job. We use the Scala programming language, developed by EPFL prof. Martin Odersky and used by tech. companies such as LinkedIn, Foursquare for their big data applications.
IPRoduct: A database of patent-product pairs 7
We are currently developing the crawler and the parser
2016/09 2017/09
Phase I(own funds)
Phase II(NSF + own funds)
Phase III(looking for sponsor-s)
Objectives Assessing feasibility Software development Small-scale database Scoping
Objectives Scaling up UKIPO, EPO data Trademark data Price data etc.
Objectives Manual collection Early assessment
Public release of code and data
Data and code sharing policywill depend on sponsors
IPRoduct: A database of patent-product pairs 8
Overview of the Phase I database
About 30 companies associated with 1,000 products and 3,000 U.S. patent documents.
In fields as diverse as cosmetics, food and beverages, computer software, telecommunications, medical devices, consumer goods, pharmaceuticals and building material.
3 Median number of patents per product(average of 7.74)
Number of patents per product
323 products are associated with only 1 U.S. patent.
Can serve as a proxy for R&D investments, complexity of technology, evidence of patent thickets, etc.
2 Median number of products per patent(average of 3.43)
Number of products per patent
1320 patents are associated with just one product.
Can be used to measure product similarity, patent importance, etc.
13 Mean age of patents associated with a product, in years
Proportion of products by average patent age
About 1% of products correspond to patents filed on average less than 3 years ago.
May be indicative of the speed of technological obsolescence.
9 Mean number of countries for whichpatent protection was sought
Proportion of patents with a family member in the jurisdiction
45% of inventions are protected at the USPTO, the EPO and the JPO.
Witnesses the high economic value of patents associated with a product.
9 Mean number of countries for whichpatent protection was sought
8 Number of years between the oldest andthe most recent patent in a product
Maximum time length between two patents linked to the same product
About 40% of products have patents filed in a time interval of maximum 2,000 days (5.5 years).
Indicative of how long a company has been developing the product, and may inform about follow-on innovation strategies.
22%
Proportion of families linked to a product
Proportion of families linked to a product
Company #1 has 11 families, and 73% of these families are linked to a product.
Families not linked to a product may signal failed commerci-alization attempts, FTO patents, etc.
The Project in BriefStylized Facts
Policy Relevance
IPRoduct: A database of patent-product pairs 18
Budget justification is an important policy aspect
The IPRoduct database will provide a concrete manifestation of S&I and the role of government funding, in a systematic manner.
Can also be used to identify case studies, in order to increase public engagement with science.
Patents Papers Grants
Procurement
UniversitiesPROs
IPRoduct: A database of patent-product pairs 19
But policy relevance goes beyond budget justification
Provides a better understanding of how IP impacts the economy.
Opens many research questions in many different disciplines (e.g., economics, management, law, marketing, science of science policy). For instance, it may provide new insights on the private returns to patenting or the value of patents.
Novel industry-level S&T indicators– Time to market: average patent age– Patent density: average number of patents per product
Also directly serves one purpose of the America Invents Act, namely to increase transparency of the patent system.
Thank you!
Gaétan de RassenfosseEcole polytechnique fédérale de Lausanne