This document discusses using machine learning to better analyze and monitor patent landscapes, specifically for emerging technologies related to Industry 4.0 and Factory 4.0. It provides an example analysis of the 3D printing patent landscape. Supervised learning can automatically categorize patents into user-defined categories and continuously learn from expert feedback to improve categorization over time, allowing for dynamic monitoring of changing technologies. This approach provides a more precise way to track patent data in fields that are poorly defined or evolving rapidly compared to traditional classification methods.
ICIC 2017: How to effectively monitor Technological Developments in IP
1. Analyzing and monitoring technological developments in the
changing patent landscape
Sample study: Factory 4.0
Contents
• ip-search Technology Fields: Definitions and analysis
• Example studies of Factory 4.0: Technology aspects in the patent landscape
• Using supervised learning to better collect, analyse and keep track of future technologies
ICIC, 23.10.2017, by ip-search
2. 2
Patent landscape analysis: What for?
DEFINITIONS
•Topic
•Package
ANALYSIS
•Basic
•Advanced
•Professional
RESULT
•Report
•Charts
•Key figures
OUR PATENT LANDSCAPE ANALYSIS
YOUR BUSINESS
DECISION
BUSINESS STRATEGY
PATENT STRATEGY
MARKET STRATEGY
PRODUCT STRATEGY
EXPERTISE
MITIGATE YOUR RISK
Smart solutions for your competitive advantage
Acquire fundamental information to make your business decision with reduced risk
ICIC, 23.10.2017, by ip-search
3. Technology fields today: Patent informations from and for a
business perspective
Consolidated analysis, human expertised full
text evaluation and data visualisation by
3. Analysis
• Visualisation
• Identification of
trends, targets
and
opportunities
• Benchmarking
• Developing
decision
ground
Supervised learning categorisation
by:
4. Active Monitoring
• Updating the selection
with high efficiency
using machine learning
and text mining
• Easy assessing and
evaluating of the data
• Re-assessing, and re-
adjusting the learned
target
ICIC, 23.10.2017, by ip-search
Examples out of 60 cutting edge technology
fields and up to 200 geographic regions,
elaborated by:
• Smart House
• Process Automation
• Fintec
• Autonomous Driving
• Li-Batteries
• 3D Printing
• Wearables
• ..
1. Define, collect and
categorise the patent
data
Added value information, such as
owner information or quality
indices delivered by:
Added value information, such as
geographic segmentation or
technology field inputs by:
• Patent owner consolidation
• Owner type identification
(University, Company)
• Legal situation (only active
patent families)
• Quality indexing over the
full database
• Geographical indication
2. Adding business-relevant
Parameters, normalize factors
4. The Patent Asset IndexTM Methodology by PatentSight:
Visualize and Compare Patent Quality on all worldwide active patent
families
20% patents worldwide
with highest rating
50% patents worldwide
with low rating
30% patents worldwide
with medium rating
ICIC, 23.10.2017, by ip-search
5. Source: Roland&Berger
Industry 4.0: from global buzz to reality
Definition of the industrial application of
Industry 4.0: «Factory 4.0»
But what about
«Factory 4.0» in
the Patent
Landscape??
USE CASE: Factory 4.0 and the patent landscape
ICIC, 23.10.2017, by ip-search
6. Patent Landscape of Industrial application of Industry 4.0: «Factory 4.0»
Technology Fields of interest: Total 1.109 Mio active patent families (Status 12/2016)
AverageQuality
CompetitiveImpact™
Forward Citation count (weighed and normalized)
Technology Relevance
Technology Fields Basic
Process Automation
Sensors
Digital Communication
Ceramics
Technology Fields
Advanced
Additive Manufacturing
Advanced Manufacturing
Blockchain/Bitcoin
Predictive Maintenance
Artifical Intelligence
3D Printing
IoT Smart House
IoT Smart City
Autonomous Driving
Robotics
Nanomaterials
Carbon&Graphene
Factory 4.0 Patents: From basic to advanced technology,
combination and integration is key
Question: What is the quality level of patents in the technological area of Factory 4.0
ICIC, 23.10.2017, by ip-search
3D-Printing
to advanced technologies
Autonomous Driving
IoT Smart House / IoT Smart City
Add.Manufact Adv.Manufact
Robotic
Art.Intelligence
Carbon&Graphene
Nanomaterial Bitcoin
and further to combined and
integrated technologies
Robotic&ArtIntelligence
Sensors&ArtIntelligence
Sensors&Robotic
DigiCom&Robotic
DigiCom&Sensor
Increasing quality
due to forward citations:
From basic technologies
Digital Communication
Sensors
CeramicsProcess Automation
Low quality
patents
Top 10%
Worldclass
patents
7. Digital Revolution (esp. in the Industry 4.0 Factory area): Its not about digital or digital
communication, but about the clever combination of technologies. Even further, the more and the better
technologies are combined, the higher the patent quality and the higher the technological maturity
Problem for patent analysts: How to collect, analyse and keep track of the patent collections of
these all technologies, combinations, changes and developments???
Factory 4.0: Assumption and Conclusion
ICIC, 23.10.2017, by ip-search
8. Worlds Average
Patent Landscape of Industrial application of Industry 4.0: «Factory 4.0»
Technology Fields of interest: Total 1.109 Mio active patent families (Status 12/2016)
AverageQuality
CompetitiveImpact™
Forward Citation count (weighed and normalized)
Technology Relevance
Technology Fields Basic
Process Automation
Sensors
Digital Communication
Ceramics
Technology Fields
Advanced
Additive Manufacturing
Advanced Manufacturing
Blockchain/Bitcoin
Predictive Maintenance
Artifical Intelligence
3D Printing
IoT Smart House
IoT Smart City
Autonomous Driving
Robotics
Nanomaterials
Carbon&Graphene
and further to cross sector
overlapping technologies
Robotic&ArtIntelligence
Sensors&ArtIntelligence
Sensors&Robotic
DigiCom&Robotic
DigiCom&Sensor
Increasing quality
due to forward citations:
From basic technologies
Digital Communication
Sensors
CeramicsProcess Automation
Factory 4.0 In-Depth Analysis: Demo Case 3D-Printing
3D-Printing
to advanced technologies
Autonomous Driving
IoT Smart House / IoT Smart City
Add.Manufact Adv.Manufact
Robotic
Art.Intelligence
Carbon&Graphene
Nanomaterial Bitcoin
Next step: In-Depth Analysis
of 3D-Printing as a demo case:
ICIC, 23.10.2017, by ip-search
9. Demo Case 3D-Printing: Categorization of technology
3D Printing Technologies:
• SLS/SLM Selective Laser Sintering (Powder/Laser Sintering
or Melting)
• SLA/DLP Stereolithography (Photocuring of Photopolymers
in Liquids)
• FDM Fused Deposition Modeling (Meltable Polymers,
Extrusion Nozzles)
• Powder/Binder-InkJet or Builder/Liquid Solidifying
• Specialty Technoloy or Application: Bioprinting or
Tissue/Organs etc.
Collected but not specifically classified: Generic 3D Patents, 3D but
Technology-unspecific materials, Digital and Software related patents, End-
products (& some noise)
Goal: Define relevant technologies and identify related patents
ICIC, 23.10.2017, by ip-search
10. 12
Demo Case 3D-Printing: Activities
First patents
Early stages
First success & scepticism
Boom phase
Question: How active are players and since when
3D printing technologies:
• SLS Selective Laser Sintering
• SLA Stereolithographie & DLP
• FDM Fused Deposition Modeling
• Powder-Binder Printing
• Bioprinting
ICIC, 23.10.2017, by ip-search
11. Demo Case 3D-Printing: Trends in Technologies
SLS
SLA
FDM
Powder-Binder
Bioprinting
Question: What kind of trends can be seen
Boom, new technology
or just a divers application?
High productivity
New successful? variants (DLP,...)
Widespread use
ICIC, 23.10.2017, by ip-search
Proper defined
technology field?
12. Demo Case 3D-Printing: Inventive countries*
SLS
SLA
FDM
Powder-Binder
Bioprinting
Question: Where are players active
* Calculation is based on inventor addresses
Technological fields:
ICIC, 23.10.2017, by ip-search
13. Demo Case 3D-Printing: Technology share of top 15
companies*
SLS/SLM
SLA/DLP
FDM
Powder-Binder
Bioprinting
Question: Who is strong in which technology
Printer
Manufacturer
Materials
manufacturer
ICIC, 23.10.2017, by ip-search
* Analysis of the entire field, calculation based on Patent Asset Index ™
14. 0,0
2,0
4,0
6,0
8,0
10,0
12,0
14,0
16,0
18,0
20,0
0 50 100 150 200 250
Stratasys
3D Systems
BASF
Electro Optical Systems
United Technologies
Evonik
HP Inc.
Carbon3D
voxeljet
Renishaw
ExOne
MarkForged
Bego
Organovo
SLM Solutions Group
16
Note: Increasing translucency of bubbles means they mark an earlier point in time.
The development over time is shown for Reporting Date 2011 to 2017.
Bubble area = Total Strength Patent Asset Index™
Sorted by PAI
AverageQuality
CompetitiveImpact™
Quantity
Portfolio Size
Demo Case 3D-Printing: Performance of Top Players
Question: How do top players develop over time
Development over the
years 2000-2017
ICIC, 23.10.2017, by ip-search
15. Conclusion Part 1:
• Collecting patents “classically” in combination with business relevant information leads to
valuable insights, …BUT.
• Badly defined, changing or dynamic technology fields make a classical approach difficult and
rather impossible to maintain and monitor regularly.
• To increase precision and to monitor such fields, we need other tools in addition to what we
have. But while all kinds of semantic or automatic tools were introduced since years, only very
few seem to work properly towards a real benefit in the patent environment.
Supervised learning engines, properly integrated and made for patents can be a game
changer
17
ICIC, 23.10.2017, by ip-search
16. • is a machine-learning based document classification software
• automatically classifies documents into customer-specific categories
• continuously learns from and imitates the behavior of IP professionals
Our approach and partner:
ICIC, 23.10.2017, by ip-search and averbis
17. Define Categories1
Provide Examples & Train2
Let the System Categorize
Documents
3
Review Results4
Active
Learning
GO
Procedure in a Nutshell
ICIC, 23.10.2017, by ip-search and averbis
18. Step 1: Define Categories
ICIC, 23.10.2017, by ip-search and averbis
19. Step 2: Provide Examples as Training Material
ICIC, 23.10.2017, by ip-search and averbis
20. Step 3: Train the System
ICIC, 23.10.2017, by ip-search and averbis
22. Step 5: Active Learning by Re-Training
ICIC, 23.10.2017, by ip-search and averbis
23. 25
Step 6: Reviewing Results and further, now
corrected and fine-tuned analysis
Averbis Internally: Statistical Analysis Externally with ip-search: Fulltext analysis or
advanced analysis in Patentsight
or export to various other tool
ICIC, 23.10.2017, by ip-search and averbis
24. Define Categories1
Provide Examples & Train2
Let the System Categorize
Documents
3
Reviewed or Unreviewed
Results
4 Regular
export to
customer
New Prototype: Active Monitoring
Feeding in
new
patents
regularly
via API etc.
ICIC, 23.10.2017, by ip-search and averbis
25. 27
Summary and Conclusions
• Actively monitoring technology field using supervised machine
learning might be the key to collect and maintain (patent) data
collections, proper, flexibel and up to date.
• Machine learning and (re-)training in the patent world is becoming
much easier, more reliable and flexible, even with few training
examples, therefore quick to adapt and refocus.
• Technology fields are valuable patent collections with multiple usage
benefits (if they are collected properly).
• Combination of data, normalized, weighted and enriched with
business information is the base for competitive advantage analysis.
• For a most comprehensive patent analysis many factors are to be
considered and experts should be consulted.
ICIC, 23.10.2017, by ip-search and averbis
26. Dr. Jochen Spuck
• Chemist, Expert in polymer chemistry
• Head of Product Development at ip-
search
• Patent Professional at the Swiss
Federal Institute of Intellectual
Property since 10 years
Dr. Kornél Markó
• Computer Scientist, Natural Language
Processing
• Co-Founder of Averbis GmbH, 2007
Questions?
ICIC, 23.10.2017, by ip-search and averbis