Transcript A szemantikus web keres*gépei

FUTURE OF MEDLINE
AND SEMANTIC WEB ENGINES
DR. GEGES JÓZSEF
OVIDIUS INFORMATION SERVICES LTD.,
WEST SUSSEX U.K.
[email protected]
1
ANSWER THE QUESTIONS
How many kind of MEDLINE you may know?
Which one do your prefer?
How many search field you use in general?
Comment the „output” form
(did you solve problem or looked for docs only)
*
2
PRINCIPLE
„ I have a dream for the Web in which
computers become capable of analyzing all
the data in the Web”
Tim Berners-Lee, 1999
3
4
WEB 3.0 CHALLENGE
(Hyper Text Markup Language)
–
(extensible)
(Resource Description Framework)
–
???
Ontology*
Know texts
LEARN TEXT - INTELLIGENT
Read – write
META SEARCH
KNOWLEDGE WEB 3.0
6
WEB 3.0 REQUIRES
•
•
•
•
•
SOLVING PROBLEM
CLEAR HITS
RIGHT FORM OF DATA
EASY TO USE
CLUSTERING
VISUALISATION
•
•
•
•
CLEAR FORM
EXACT WORDS
SIMPLE COMPETENCY
CLEAR FORM AND
STUCTURE OF DATABASES
7
SOLUTIONS
GRAFICAL CLUSTER
FAQ
PROFESSIONAL DATA
www.google.com
www.carrot2.org
www.touchgraph.com
www.ask.com
www.trueknowledge.com
www.yebol.com
SEMANTIC SEARCHING
8
MEDLINE & WEB 3.0
• EVERYBODY CAN UNDERSTAND „MESH”
• NATIONAL LIBRARY OF MEDICINE HAS THE BIGGEST WELL
STRUCTURED DATABASES
• EASY TO CONTROL SEARCH
• EVERYBODY HAS HIS OWN SEARCHING STRATEGY
9
MEDLINE VARIATIONS
1.
2.
3.
4.
5.
LIST OF HITS AND PRIORITIES
RANKING BY PRINCIPLE
RANKING BY RELATIONS AND CONNECTIONS
USER INTERFACE
HOW TO ACCESS DATA
WE KNOW AND USE ABUT 30 MEDLINE VERSIONS
10
PUBMED
12
PUBMED CHANGES
13
MEDLINE HITS AND RANKING
SEMMED
REFMED*
QUERTLE
MEDLINERANKER
MISEARCH*
HAKIA
SEMANTICMED 
MSCANNER
ETBLAST
PUBFOCUS
TWEASE
2010 RANKING BY FEADBACKS - QUALITY
2009 BY PRINCIPLES
2009 CUSTER ADDED RANKING
2009 FEEDBACKS
2008 SEMANTIC SEARCH DEVELOPED
2008 TECHNOLOGY OF COGNITIONS
2008 MULTI LEVEL CLUSTERING
2007 RELATED SEARCHES
2006 BY IMPACT FACTORS
2005 AMENDED MESH
http://skr3.nlm.nih.gov/SemMedDemo/
14
SEMANTIC WEB
„FIND WHAT I MEAN NOT WHAT I TYPE”
GO ENGINES
- GERMAN BASED DEVELOPMENT (2010)
- MESH plus ONTOLOGY
- WIDELY USED ALGORITHM
- USERFRIENDLY, CLUSTERING VISUALIZATION
http://www.gopubmed.org
15
SEMANTIC WEB
„FIND WHAT I MEAN NOT WHAT I TYPE”
COGNITION
- 24 years of development
- multidisciplinary
- TESTED ON MEDLINE and WIKIPEDIA
- Poor visualisations
http://www.cognition.com
16
SEMANTIC WEB
„FIND WHAT I MEAN NOT WHAT I TYPE”
HAKIA
- the best one, but limited
- multilanguage
- runs generally 10 „cluster” at the same time
- sophisticated alternative keywords, like: cure, treat,
therapy, look after
http://hakia.com/
17
SEMANTIC WEB
„FIND WHAT I MEAN NOT WHAT I TYPE”
DEEPDYVE
- DeepWeb and document searches
- sophisticated clustering
- very long search sentences
- little bit complicated list of hits
http://www.deepdyve.com/
19
WHAT ELSE OF SEMANTIC WEB BUT NOT
MEDLINE RELATED
SWSE – the best technology
KOSMIX – Time Warner spent $20M
EXALEAD – focused on picture
LEXXE – Q&A type (2011 Sept. 27 restarting)
POWERSET – bought by Microsoft – BING
SWOOGLE – retired
20
TWO WAYS TO FOLLOW
„The WellPoint application will combine data from three sources: a patient's
chart and electronic records that a doctor or hospital has, the insurance
company's history of medicines and treatments, and Watson's huge library of
textbooks, medical journals and databases.”
21
22