Finding trends in large scale document sets

Treparel Dr. Anton Heijs
Delftechpark 26 CEO/CTO
anton@treparel.com
2628 XH Delft
The Netherlands
September 24, 2012
www.treparel.com

Analysing large patent portfolios: Nothing
remains uncovered

Agenda
• Introduction Treparel &
KMX Patent Analytics
• How to deal with Big Data
in patents?
• Landscape analysis of
large patent sets
• Use Cases: SWOT analysis

Fig 1: patent landscape of ebook technologies

Treparel KMX – All rights reserved 2012 www.treparel.com 2

About Treparel
• Treparel is an innovative technology solution provider of
– Big Data Text Analytics and Visualization technology
– Patent Analytics solutions
• KMX is an integrated data analysis toolset which provides
– Fast and accurate insights in large unstructured document sets
to allow companies to make better informed decisions.
• KMX software platform
– Strong focus on R&D with university ecosystem
– Over 30 man years of software-development
– Used by knowledge driven organizations in technology, chemical,
and life sciences
• Based in Delft, The Netherlands since 2006.

IP landscape trend example: 1970-2007

Alstom GE

Dong Feng Siemens 4

The importance of analytics in IP

Research Discovery Development Market Life Cycle
Launch Management

Market, Legal and Competitive Analysis
Patent Analysis
Research Analysis

$

Investment
Decreasing return Increasing Return
years

Economic changes in the last decades
• Globalization :
– More companies, increased competition, lower margins
– Innovation for shorter product life cycles
– Manufacturing has become a commodity by outsourcing to
low wage countries
• The cost of R&D increased but competition leads to price
erosion
• Return on R&D investment drives importance for value
creation from IP
• Competitive edge of companies shifts from
production-based to knowledge-based


The Intellectual Economy
Past : Selling products finances R&D
Research , Development , Marketing , Life Cycle
Discovery Manufacturing Sales Management

Revenues
Return on investments
Today: value creation and revenue generation from IP (to finance R&D)

Research , Development , Marketing , Life Cycle
Discovery Manufacturing Sales Management

Innovations

IP Licensing

Revenues Revenues

Return on investments
Treparel KMX – All rights reserved 2012 www.treparel.com
7

Adapting to economic developments

The New Reality Value from IP portfolio
• IP strategy over time • Protect market share
– Increase the value of the IP portfolio • Save cost for royalties via cross
– Maximize the ROI from R&D licensing
• License management • Generate income from royalties
– analyze the impact of economic developments • Benefit from License Out from
and its effect for the IP strategy joint ventures or spin-offs

Drivers
• Number of Patent filings is growing
• Growing need to drive revenues from
licensing to fund R&D investment


Trends and implications for the future
• Economies depend stronger on each other
• Innovation is a driver of economic growth
• Globalization generates many patents from China
and South Korea
• The number and complexity of patent filings is
growing

Growing need for fast and accurate analysis of
large sets of patents


Getting more insight using less experts
Big Data Paradox:
• Limited (human) resources available for in-depth analysis
• Growing need for data driven decisions

Growing Data, Faster Insights: Big Data Analytics

Velocity Volume

Variety Complexity


The big data paradox ; more data but less
knowledge ‘Information Gap’

10

Available data
8 Information Gap
6 Data driven decisions

4

2 Available experts for supporting decision making

0

1990 1995 2000 2005 2010 2015 2020


Information democracy:
Information Creators and Consumers
• Creators
– Defines & prepopulate Analysis Pipelines and test it on
the data
– Deploys these pipeline using Cloud computing
• Required computing capacity can scale up with the business /
analytical needs
• Consumers
– Pre defined analytical reports
– Sharing feedback and input to the results to optimize
analysis

Sharing information empowers collaboration

The traditional IP search and analysis
Request

Results
Database Tools

Data Information Creator Request Information Information Consumer

The traditional approach:
• Each search/analysis request is focussed on a specific question from one user
• When the number of request increases this requires more human searchers
• When the searches involve analysis of more patents this requires more time
• Very specific searches can not be automated
• Analysis of large documents sets can be automated – which is an opportunity to
analyse more and become more competitive


Information democracy:
Proactive analysis to search and analysis
Patent Analyst
Database

Research
Database Running
Business
Analytics User
Marketing Pipeline
Database

Data Information Creators Push Information Information Consumers

Liberate the Information Search, facilitate the discovery process:
• The knowledge Creator:
• defines the analysis pipeline and test it on the data
• deploys the analyses using cloud computing resources
• Direct access for information consumers for in depth analyses

Liberate more information using new
technologies
• Enable a small group of experts with tools to set-up IP
analysis pipelines
– Extend the search on request approach with a pro-
active analysis approach on large document sets
– Use analysis pipelines to auto generate visualizations
in a browser
– Invest in new technology coming from Big Data
Analytics
• Give a large group of users access to the internal
webpages providing them with rich statistical
information and interactive visualizations


Combining proactive analysis with traditional
IP search and discovery
Personal Request
Tools

Tailor made Results
Database

Patent Business
Database User

Running Exchange
Information
Research
Database
Analytics
Pipeline Analyst

Marketing
Database

Data Information Creators Information Information Consumers

Performing small to large scale SWOT
analysis
SWOT analysis example
Patent
Database Queries • What are most important
patents?
• Who owns them?
• What is growth of
patents by:
• Technology?
• Owner?
• Country?
• Year?

5000 patents 1000 patents 500 patents

Business Overview
User and details
Ranking Filtering


Auto reporting & analysis for multiple users

• Reporting of aggregated
results:
– Pie & bar charts
• Providing overview of the
subject:
– landscape visualization
• Enabling rich interaction

Page 18

Use Case1: SWOT analysis of Ebooks
• Perform proactive SWOT analysis of ebooks market
Amazon Kindle – Apple – Samsung/Google and other players

• Who owns what?
• What can we learn from competitive technology landscape?
• Why?
• Determine a company/technology position and opportunities
• We do this in KMX by:
1. Query to get patents on electronic paper technology
2. Landscape analysis
3. Classification/Ranking
4. Filter and select subset
5. Iterate step 1-to-5

Treparel KMX – All rights reserved 2012 Fig 2: Overview landscape visualization of 4257 patens 19
19

Analysis of ebook technology

Fig 3: Overview landscape visualization of 4257 patents

Use document classification to rank the
patents
Purple = most important patents
Red = least relevant patents

Fig 4: Ranked patents using a classifier for ebook technology (In purple the selection of relevant
patents for deeper analysis)

Drill deeper in the data to learn more

Fig 5: Landscape visualization going from 4257 to 1049 to 369 patents

After removing the irrelevant patents we use filtering to
determine:
• Who are the important players (assignees, inventors)?
• Where are the important patents filed (countries)?
• What is the trend over time (growth of patents over the years)?


Define the relevant set of patents to identify
your strengths & opportunities

Purple = most important patents
Red = least relevant patents
Fig 6: Landscape visualization of 369 most important patents in ebook technology
Treparel KMX – All rights reserved 2012 23

The role of language:
Clustering of Patents in Chinese text

Fig 7: Patent landscape visualization using the chinese or englisch text

Use Case 2: SWOT Technical & Biological patents

• Perform SWOT analysis in a converging market: analyze claims with mixed
technologies
• Who owns what?
• How does the technology landscape looks like?
• Why ?
• Determine a company/technology position and opportunities
• We do this in KMX by:
1. Query to get patents on mechanical/electronic/optics mixed with
biological technology
2. Landscape analysis
3. Classification/Ranking
4. Filter and select subset
5. Iterate step 1-to-5

Treparel KMX – All rights reserved 2012 Fig 8: Landscape visualization of 10920 patents 25

Use Case: SWOT analysis : patents covering
multiple technology areas (engineering & biology)

• Patents become more
complex to analyse

• Examples:
• More detailed claims
• Mixed technologies in
the claims

• Obtaining an landscape
overview is then key

• Analysis from the users
perspective is essential
(classification/ranking)

Fig 9: Landscape visualization of 10920 patents


Patents from total set with biological focus
• Using the text from the
title/abstract and claims

• landscape analysis
provides overview

• Using document
classification to
determine sub clusters

• Using classification and
ranking to determine
the most relevant
documents from the
users perspective

Fig 10: Landscape visualization of 134 patens

Key takeaways
 Global economy demands value generation from an IP strategy
 Big Data Paradox:
• Limited (human) resources available for in-depth analysis
• Growing need for data driven decisions

 The information creators (patent searchers) focus on
• Providing proactive information on generic analysis tasks
• Perform specific analysis for single user request

 The information consumers (patent council)
• Get knowledge from automated analysis with
interactive capabilities
• Obtain SWOT analysis knowledge of competitors
• Built and optimize patent strategies


Finding trends in large scale document sets

Recommandé

Recommandé

Contenu connexe

Similaire à Finding trends in large scale document sets

Similaire à Finding trends in large scale document sets (20)

Dernier

Dernier (20)

Finding trends in large scale document sets