Transcript eClassifier / MindMap

Almaden Services Research
Intellectual Property Analytics
Turning Unstructured Information Into Value
Jeffrey T. Kreulen, Ph.D.
[email protected]
W. Scott Spangler
[email protected]
© 2008 IBM Corporation
Almaden Services Research
Innovation is About More Than Building a Better Mouse Trap …
© 2008 IBM Corporation
Almaden Services Research
It’s About Beating The Competition …
© 2008 IBM Corporation
Almaden Services Research
From Transaction To Interaction Business Intelligence
Execute ControlDirect
Suppliers
Context
Business Competencies
 Classic BI focuses on transactions at the
boundary of the enterprise
The Enterprise
Control
Direct
Business Competencies
Execute
 Contextual information and its derived
intelligence has to be integrated into the
overall picture
Execute ControlDirect
Customers
ExecuteControlDirect
Business Partners
Business Competencies
 Transactions are evolving to include
more of the enterprise eco-system and a
richer set of the life-cycle of interactions
Business Competencies
 Incorporation of unstructured information
into individual and collection level
analytics
© 2008 IBM Corporation
Almaden Services Research
Leveraging Interactions in the Enterprise Eco-System
CEO Survey: Sources of New Ideas and Innovation
CRM / Call Centers
Jams
External
Internal
Employees (general population)
Business partners
Customers
Sales or service units
Classic BI
Consultants
R&D (internal)
Competitors
Other
IP, Web, …
Think tanks
Associations, trade groups, conference boards
Academia
45%
35%
25%
Procter & Gamble has set itself a goal
of getting half its new product ideas
from outside the company by 2010.
15%
Internet, blogs, bulletin boards
5%
5%
15%
25%
35%
45%
“We have...today a lot more capability and
innovation in the [competitive] marketplace...
than we [could] try to create on our own.”
IBM Institute for Business Value, CEO Study 2006
© 2008 IBM Corporation
Almaden Services Research
The Analytical Approach
Understand the Business Objective
Identify Information Sources
Explore
- Multiple data
source collections
- On-Topic data search
- Nearest neighbor search
-Intermediate
analytics results
Understand
- Taxonomy Generation
- Clustering
- Classification
- Dictionary
- Synonyms
- Editing
- Refinement
Analyze
- Trending
- Network analysis
- Co-occurrence
- Scatter plot
- Graphing/Reporting
- Visualizations
-...
© 2008 IBM Corporation
Almaden Services Research
Business Analytical Solutions
Almaden
Patent
Analytical
Workbench
(SIMPLE)
Corporate
Brand and
Reputation
Analysis
(COBRA)
Service
Delivery
Insight
(SDI)
Call Center Analysis
Business Information Services On the Network (BISON)
An SOA implementation of information analytical capabilities used to enable solution developement
Business Insights Workbench (BIW)
A comprehensive platform for structured and unstructured information analytics
© 2008 IBM Corporation
Almaden Services Research
Business Objectives of Patent Analytics
 Prior Art Search
 Strategic Analysis
– Competitive Landscape
– Technology Landscape
– People Landscape
– Mergers & Acquisitions
 Portfolio Management
– Partnering and Licensing
– Defensive
– Valuation
© 2008 IBM Corporation
Almaden Services Research
IBM Patent Portfolio Since 1994
A simple taxonomy of IBM’s patent portfolio since 1994 with counts and
sorted by recency. The portfolio is migrating to the categories at the top (see
summary trend lines).
© 2008 IBM Corporation
Almaden Services Research
Using patent data from last 18 years, we compared the most
relevant concepts to identify emerging patterns
Total # of
patents
Pfizer
Merck
BMS
J&J
AstraZeneca
Novartis
Amgen
 By comparing the most relevant
concepts in patent data, we
observed patterns emerging.
 Genentech is staking out white
space in the areas not covered by
the other major pharmaceuticals.
 Looking at US patent data for the
last 18 years shows how
pharmaceutical companies are
positioning themselves in the market.
 Leading companies like Pfizer,
AstraZeneca, and Amgen are
increasing their patent activity while
other companies are decreasing.
Merck
Bristol Myers
Squibb
Novartis
Johnson & Johnson
© 2008 IBM Corporation
Almaden Services Research
SIMPLE
Strategic Information Mining Platform
 Full text search on all US
patents and applications
 Nearest Neighbor search from
a list of input patents or
document text
 Claims originality Analysis
 Divestiture Impact Analysis
 Chemical Structure Searches
 Patent Clustering
 View IBM patent status
information from Dossier
 Create and save patent
“projects” and export reports
© 2008 IBM Corporation
Almaden Services Research
BISON - Component Architecture for SIMPLE
Patent
Search
UI
Claims
Originality
UI
Nearest
Neighbor
UI
Patent
Clustering
UI
Divestiture
UI
Dossier
Search
UI
Citation
analysis
UI
BIW
analytics
application
UI
Chemical
Search
UI
Tapestry Discovery Framework
ssc Discovery
Framework
AXIS Web Services Client
SSC Commons
AXIS Web
Services
Client
User Interface layer
WebSphere 6.1 J2EE Container
User Interfaces
Divestiture
Patent
Search
Nearest
Neighbor
Dossier
Search
Analytics Services Layer
Claims
Origniality
Analytics Services
Patent
Clustering
Chemical
Search
BIW
Analytics
Framework
Citation
analysis
WEB Services BUS
Customer
Hosted
Databases
DAS (Data access Services)
Solr
Grants
index
Solr
Applications
index
Solr
Dossier
index
Indices
Indexing Layer
AXIS 1.x based Web Service Bus
Annotation
Process
Data Storage Layer
Data Replication
Master
IP Database
Schema
Published
IP Database
Schema
BlueGene L
DOSSIER
Database
Data Sources
GETL
Patent XML DataFeeds
Extraction transformation
and Load
Data Sources
© 2008 IBM Corporation
Almaden Services Research
SIMPLE Patent Searching
Full-text and fielded
searching for US, EP
and WIPO patents.
© 2008 IBM Corporation
Almaden Services Research
Nearest Neighbor Searching
Novel collection based
prior art searching and
similarity metrics.
© 2008 IBM Corporation
Almaden Services Research
Claims Originality
Identification of novel
language in claims.
Useful in identifying
emergent topics and
seminal patents.
© 2008 IBM Corporation
Almaden Services Research
Divestiture
A portfolio
management tool
useful in determining
relative value.
© 2008 IBM Corporation
Almaden Services Research
Patent Clustering
The ability to
automatically group a
collection of patents into
semantically similar
groups.
© 2008 IBM Corporation
Almaden Services Research
Chemical Search
We have identified
occurrences of chemical
compounds,
disambiguated by
structure (using InChi
and SMILES) and have
a chemical similarity
search engine.
Demo at http://chemsearch.almaden.ibm.com
© 2008 IBM Corporation
Almaden Services Research
And of course, this doesn’t just work for patents…
© 2008 IBM Corporation
Almaden Services Research
COBRA – Corporate Brand and Reputation Analysis
Themes are configured for brands, competitors, people and other
topics of interest.
The system filters blogs, boards, and
online news to identify the snippets
which contains the information of
interest
Using orthogonal filtering techniques
we can get an accuracy rate of 95%.
Even with this, the users still need to
look at the document for context
User define folders provide the user to
organized the alerts into their own
taxonomy.
• We use multiple techniques to identify documents
of interest for alerting and analysis.
• We are creating industry templates.
© 2008 IBM Corporation
Almaden Services Research
COBRA – Corporate Brand and Reputation Analysis
Sentiment Analysis is combined with Themes
Just know that a number of posting
are negative tell only part of the
story. Trends help determine if this
is something to be concerned about
or “old news”
There were 20 blog postings
which were identified with
negative sentiment.
© 2008 IBM Corporation
Almaden Services Research
COBRA – Corporate Brand and Reputation Analysis
Interactive Analytical Dashboard
© 2008 IBM Corporation
Almaden Services Research
COBRA – Corporate Brand and Reputation Analysis
We look for interesting correlations
1
3
2
4
5
© 2008 IBM Corporation
Almaden Services Research
How do
Jams work?
© 2008 IBM Corporation
Almaden Services Research
Service Delivery Insight
 Call Center mining and analysis
© 2008 IBM Corporation
Almaden Services Research
http://www.miningthetalk.com
Brazen
Plug!
© 2008 IBM Corporation