![Page 1: Exploring Automated Patent Search with KNIMEfiles.meetup.com/19483337/KNIME_Den_Haag_Alexander... · 2016-03-14 · Exploring Automated Patent Search with KNIME Possibilities, Limits,](https://reader033.vdocument.in/reader033/viewer/2022053000/5f0473f37e708231d40e0a88/html5/thumbnails/1.jpg)
Exploring Automated Patent
Search with KNIME Possibilities, Limits, Future
Alexander Klenner-Bajaja, PhD
![Page 2: Exploring Automated Patent Search with KNIMEfiles.meetup.com/19483337/KNIME_Den_Haag_Alexander... · 2016-03-14 · Exploring Automated Patent Search with KNIME Possibilities, Limits,](https://reader033.vdocument.in/reader033/viewer/2022053000/5f0473f37e708231d40e0a88/html5/thumbnails/2.jpg)
European Patent Office
2
Offices: Berlin, Vienna, Munich, The Hague (Rijswijk), Brussels
Staff: over 7000
Mission: Granting European Patents
New building in Rijswijk,
finished approx. 2018
![Page 3: Exploring Automated Patent Search with KNIMEfiles.meetup.com/19483337/KNIME_Den_Haag_Alexander... · 2016-03-14 · Exploring Automated Patent Search with KNIME Possibilities, Limits,](https://reader033.vdocument.in/reader033/viewer/2022053000/5f0473f37e708231d40e0a88/html5/thumbnails/3.jpg)
Why Search – European Patent Convention
3
Information Management`s
Task: Support Search
![Page 4: Exploring Automated Patent Search with KNIMEfiles.meetup.com/19483337/KNIME_Den_Haag_Alexander... · 2016-03-14 · Exploring Automated Patent Search with KNIME Possibilities, Limits,](https://reader033.vdocument.in/reader033/viewer/2022053000/5f0473f37e708231d40e0a88/html5/thumbnails/4.jpg)
Patent Search – Filing numbers increase
4
2005 -----------------------------------------------------------2015
![Page 5: Exploring Automated Patent Search with KNIMEfiles.meetup.com/19483337/KNIME_Den_Haag_Alexander... · 2016-03-14 · Exploring Automated Patent Search with KNIME Possibilities, Limits,](https://reader033.vdocument.in/reader033/viewer/2022053000/5f0473f37e708231d40e0a88/html5/thumbnails/5.jpg)
Introduction – What do we want, where are we?
5
Application Query Formulation
PriorArt
Search
relevant documents
![Page 6: Exploring Automated Patent Search with KNIMEfiles.meetup.com/19483337/KNIME_Den_Haag_Alexander... · 2016-03-14 · Exploring Automated Patent Search with KNIME Possibilities, Limits,](https://reader033.vdocument.in/reader033/viewer/2022053000/5f0473f37e708231d40e0a88/html5/thumbnails/6.jpg)
MetaData
Neo4J
The EPO`s Tool landscape
6
Translation
API Boolean Search
Engine
Lucene Search
Engine
Fulltext
MongoDB GoldStandard
SQL
Search
Evaluation
ImageDB
Concept
Extraction
![Page 7: Exploring Automated Patent Search with KNIMEfiles.meetup.com/19483337/KNIME_Den_Haag_Alexander... · 2016-03-14 · Exploring Automated Patent Search with KNIME Possibilities, Limits,](https://reader033.vdocument.in/reader033/viewer/2022053000/5f0473f37e708231d40e0a88/html5/thumbnails/7.jpg)
MetaData
Neo4J
The EPO`s Tool landscape
7
Translation
API Boolean Search
Engine
Lucene Search
Engine
Fulltext
MongoDB
GoldStandard
SQL
Search
Evaluation
ImageDB
Concept
Extraction
![Page 8: Exploring Automated Patent Search with KNIMEfiles.meetup.com/19483337/KNIME_Den_Haag_Alexander... · 2016-03-14 · Exploring Automated Patent Search with KNIME Possibilities, Limits,](https://reader033.vdocument.in/reader033/viewer/2022053000/5f0473f37e708231d40e0a88/html5/thumbnails/8.jpg)
MetaData
Neo4J
The EPO`s Tool landscape
8
Translation
API Boolean Search
Engine
Lucene Search
Engine
Fulltext
MongoDB
GoldStandard
SQL
Search
Evaluation
ImageDB
Concept
Extraction
![Page 9: Exploring Automated Patent Search with KNIMEfiles.meetup.com/19483337/KNIME_Den_Haag_Alexander... · 2016-03-14 · Exploring Automated Patent Search with KNIME Possibilities, Limits,](https://reader033.vdocument.in/reader033/viewer/2022053000/5f0473f37e708231d40e0a88/html5/thumbnails/9.jpg)
MetaData
Neo4J
The EPO`s Tool landscape
9
Translation
API Boolean Search
Engine
Lucene Search
Engine
Fulltext
MongoDB
GoldStandard
SQL
Search
Evaluation
ImageDB
Concept
Extraction
![Page 10: Exploring Automated Patent Search with KNIMEfiles.meetup.com/19483337/KNIME_Den_Haag_Alexander... · 2016-03-14 · Exploring Automated Patent Search with KNIME Possibilities, Limits,](https://reader033.vdocument.in/reader033/viewer/2022053000/5f0473f37e708231d40e0a88/html5/thumbnails/10.jpg)
One platform to allow rapid prototyping and
evaluation
10
![Page 11: Exploring Automated Patent Search with KNIMEfiles.meetup.com/19483337/KNIME_Den_Haag_Alexander... · 2016-03-14 · Exploring Automated Patent Search with KNIME Possibilities, Limits,](https://reader033.vdocument.in/reader033/viewer/2022053000/5f0473f37e708231d40e0a88/html5/thumbnails/11.jpg)
The current Search System
A Lucene elastic search based system, documents are returned as
ranked lists
11
Search
Spacek
Lucene query1
![Page 12: Exploring Automated Patent Search with KNIMEfiles.meetup.com/19483337/KNIME_Den_Haag_Alexander... · 2016-03-14 · Exploring Automated Patent Search with KNIME Possibilities, Limits,](https://reader033.vdocument.in/reader033/viewer/2022053000/5f0473f37e708231d40e0a88/html5/thumbnails/12.jpg)
Patent Gold Standards
We have “manually” curated search reports for about 40 million simple
patent families
The relevant documents are mentioned in the search report as either
–X(I,N),A,Y,... documents
12
median: 5 citations
in search reports
![Page 13: Exploring Automated Patent Search with KNIMEfiles.meetup.com/19483337/KNIME_Den_Haag_Alexander... · 2016-03-14 · Exploring Automated Patent Search with KNIME Possibilities, Limits,](https://reader033.vdocument.in/reader033/viewer/2022053000/5f0473f37e708231d40e0a88/html5/thumbnails/13.jpg)
Setting up a benchmarking environment
We need to move away from anecdotal evidence to statistically
meaningful facts
TAPAS
13
SEARCH
INDEX
Applications
Method 1 Method 2
MAP:0.4 MAP:0.2
Patent Corpus
* Exploiting real queries
Evaluate!
![Page 14: Exploring Automated Patent Search with KNIMEfiles.meetup.com/19483337/KNIME_Den_Haag_Alexander... · 2016-03-14 · Exploring Automated Patent Search with KNIME Possibilities, Limits,](https://reader033.vdocument.in/reader033/viewer/2022053000/5f0473f37e708231d40e0a88/html5/thumbnails/14.jpg)
Using KNIME to enhance automated search
14
Application
Query
Formulation
PriorArt
Search
relevant documents
Evaluation and Feedback
![Page 15: Exploring Automated Patent Search with KNIMEfiles.meetup.com/19483337/KNIME_Den_Haag_Alexander... · 2016-03-14 · Exploring Automated Patent Search with KNIME Possibilities, Limits,](https://reader033.vdocument.in/reader033/viewer/2022053000/5f0473f37e708231d40e0a88/html5/thumbnails/15.jpg)
Use Case: Does Translation help to find relevant
Prior Art?
15
![Page 16: Exploring Automated Patent Search with KNIMEfiles.meetup.com/19483337/KNIME_Den_Haag_Alexander... · 2016-03-14 · Exploring Automated Patent Search with KNIME Possibilities, Limits,](https://reader033.vdocument.in/reader033/viewer/2022053000/5f0473f37e708231d40e0a88/html5/thumbnails/16.jpg)
Evaluating the results – Does Translation help to
find relevant Prior Art?
16
RelevantExpected
Combinedresults
TranslatedQueries
OriginalQueries
%-relevant retrieved 100 36,92307692 27,69230769 11,53846154
0
10
20
30
40
50
60
70
80
90
100
%-r
ele
va
nt
retr
ieve
d%-relevant retrieved
![Page 17: Exploring Automated Patent Search with KNIMEfiles.meetup.com/19483337/KNIME_Den_Haag_Alexander... · 2016-03-14 · Exploring Automated Patent Search with KNIME Possibilities, Limits,](https://reader033.vdocument.in/reader033/viewer/2022053000/5f0473f37e708231d40e0a88/html5/thumbnails/17.jpg)
Patents have multi-modal information content: Images
Images
– Chemical Formulas
– Flow Diagrams
– Circuits
– Technical Drawings
17
![Page 18: Exploring Automated Patent Search with KNIMEfiles.meetup.com/19483337/KNIME_Den_Haag_Alexander... · 2016-03-14 · Exploring Automated Patent Search with KNIME Possibilities, Limits,](https://reader033.vdocument.in/reader033/viewer/2022053000/5f0473f37e708231d40e0a88/html5/thumbnails/18.jpg)
Image Search the Google way
Standard Google image search (very strong on real world images)
is currently not suited for technical drawings
Extremely complicated endeavour
18
results
![Page 19: Exploring Automated Patent Search with KNIMEfiles.meetup.com/19483337/KNIME_Den_Haag_Alexander... · 2016-03-14 · Exploring Automated Patent Search with KNIME Possibilities, Limits,](https://reader033.vdocument.in/reader033/viewer/2022053000/5f0473f37e708231d40e0a88/html5/thumbnails/19.jpg)
Use Case: Image Search
19
Search
Space
Query
State of the art
Image processing
Filtering and
Visualisation
![Page 20: Exploring Automated Patent Search with KNIMEfiles.meetup.com/19483337/KNIME_Den_Haag_Alexander... · 2016-03-14 · Exploring Automated Patent Search with KNIME Possibilities, Limits,](https://reader033.vdocument.in/reader033/viewer/2022053000/5f0473f37e708231d40e0a88/html5/thumbnails/20.jpg)
Image Search Adaptive Hierarchical Density Histogram (AHDH)
closest matchAHDH
20
New
results
Exploit distribution of image points
![Page 21: Exploring Automated Patent Search with KNIMEfiles.meetup.com/19483337/KNIME_Den_Haag_Alexander... · 2016-03-14 · Exploring Automated Patent Search with KNIME Possibilities, Limits,](https://reader033.vdocument.in/reader033/viewer/2022053000/5f0473f37e708231d40e0a88/html5/thumbnails/21.jpg)
Can we optimize our Search Tool Parameters?
21
![Page 22: Exploring Automated Patent Search with KNIMEfiles.meetup.com/19483337/KNIME_Den_Haag_Alexander... · 2016-03-14 · Exploring Automated Patent Search with KNIME Possibilities, Limits,](https://reader033.vdocument.in/reader033/viewer/2022053000/5f0473f37e708231d40e0a88/html5/thumbnails/22.jpg)
Use Case: Parameter Optimisation
22
Swarm Optimization
![Page 23: Exploring Automated Patent Search with KNIMEfiles.meetup.com/19483337/KNIME_Den_Haag_Alexander... · 2016-03-14 · Exploring Automated Patent Search with KNIME Possibilities, Limits,](https://reader033.vdocument.in/reader033/viewer/2022053000/5f0473f37e708231d40e0a88/html5/thumbnails/23.jpg)
Can we optimize our Search Tool Parameters?
23
Start
End
![Page 24: Exploring Automated Patent Search with KNIMEfiles.meetup.com/19483337/KNIME_Den_Haag_Alexander... · 2016-03-14 · Exploring Automated Patent Search with KNIME Possibilities, Limits,](https://reader033.vdocument.in/reader033/viewer/2022053000/5f0473f37e708231d40e0a88/html5/thumbnails/24.jpg)
What are the possibilities?
We were able to create a system where all components sharing the
same (table based) interface.
We make full use of many existing components and combine them
successfully with internal and external additional developments
– Text mining
– Image Similarity
– Machine Learning based document similarity
– Influence of translation to patent search
Rapid Prototyping
Evaluation and Feedback
24
![Page 25: Exploring Automated Patent Search with KNIMEfiles.meetup.com/19483337/KNIME_Den_Haag_Alexander... · 2016-03-14 · Exploring Automated Patent Search with KNIME Possibilities, Limits,](https://reader033.vdocument.in/reader033/viewer/2022053000/5f0473f37e708231d40e0a88/html5/thumbnails/25.jpg)
What are the limits?
Very practical issues like memory freezes of the eclipse Environment
Sometimes extreme over head introduction e.g. text mining nodes
(String to Document)
Not suitable from our experience for full corpus analytics (100 million
full text documents and more)
Getting lost in “yellow” nodes (ungroup, pivot, regroup, filter, missing
value,...)
25
![Page 26: Exploring Automated Patent Search with KNIMEfiles.meetup.com/19483337/KNIME_Den_Haag_Alexander... · 2016-03-14 · Exploring Automated Patent Search with KNIME Possibilities, Limits,](https://reader033.vdocument.in/reader033/viewer/2022053000/5f0473f37e708231d40e0a88/html5/thumbnails/26.jpg)
Future
Exposing stable services as web service through the KNIME server
web-interface
Exposing workflow creation to a wider range of user (right now its IM
only)
Connecting more services
Using of streaming to overcome some of the limits
26
![Page 28: Exploring Automated Patent Search with KNIMEfiles.meetup.com/19483337/KNIME_Den_Haag_Alexander... · 2016-03-14 · Exploring Automated Patent Search with KNIME Possibilities, Limits,](https://reader033.vdocument.in/reader033/viewer/2022053000/5f0473f37e708231d40e0a88/html5/thumbnails/28.jpg)
28
![Page 29: Exploring Automated Patent Search with KNIMEfiles.meetup.com/19483337/KNIME_Den_Haag_Alexander... · 2016-03-14 · Exploring Automated Patent Search with KNIME Possibilities, Limits,](https://reader033.vdocument.in/reader033/viewer/2022053000/5f0473f37e708231d40e0a88/html5/thumbnails/29.jpg)
Supplement
29