111 Blumauer - A Five-Star Rating System for Semantic Search

© Semantic Web Company - http://www.semantic-web.at/ and http://www.poolparty.biz/

A Five-Star Rating System for Semantic SearchAndreas BlumauerCEO of Semantic Web Company

http://www.semantic-web.at/

http://www.poolparty.biz/


Key questions in this talk

● Beyond precision and recall: What makes a ‘good’ search?

● Beyond document search: What other types of ‘semantic search’ exist?

● Beyond search: What other use case scenarios than ‘search’ benefit from linked data and knowledge graphs?

2




Connotation of ‘semantic’

From: Wortschatz Leipzig - http://wortschatz.uni-leipzig.de/

3



http://wortschatz.uni-leipzig.de/


Some facts and figures

Vertical search & Enterprise search:● Currently 201 ‘Semantic Search’ Startups (https://angel.co/semantic-search)● Around US$ 900 million of venture capital

Web search:● Wikipedia’s pageviews declined 21% the year after Google Knowledge Graph

was launched● Google search volume for json-ld

4



https://angel.co/semantic-search


Diverse promises made by ‘Semantic Search’ vendors

Use Weotta and get inspired.

Search is outdated. Google steers you to right section of the library, but doesn’t answer your question or compile that answer with others to help you make a decision.

Ambiance can automatically match the meaning of the page to the products that are most contextually relevant in real time.

Benefit from additional insights through interactive visualizations of reports and search results derived from your data lake

5



http://vurb.com/

http://www.weotta.com/

http://www.discoversophia.com/


So, what is ‘Semantic Search’?

It’s all about improved user experience, no matter how complex information procurement is:

I am a web shopper. I want to have nice buying experiences when using an online shop.

I am a doctor. I need a 360-degree view on my patient. I don’t have the time to search around.

6




Semantic Search is about precise contextualization (personalisation) Semantic search engines are able to link entities in a meaningful way. They extend graphs ‘on-the-fly’ by reusing already existing knowledge graphs.

“Which skateboard fits to my jacket and its manufacturer also produces snowboards?”

“Which new medical studies about allergy ABC talk also about symptoms related to my current patient?”

7




A simple semantic search engine


eyes allergy symptoms Go!

ocular allergy symptom

itchy eyes

8




A simple semantic search engine


eyes allergy symptoms Go!

ocular allergy symptom

itchy eyes

There are various semantic technologiesto produce this additionalcontextualization.

9




‘Search’ by extending graphs ‘on-the-fly’

owns fits with

fits with“Which skateboard fits with my jacket and its manufacturer also produces snowboards?”

produces

produces

Snowboardis a

Skateboardis a

10




‘Semantic Re-Engineering’ based on Text Mining

11




Semantic Authoring

“Organizations should focus on semantic consistency and performance in upstream applications and data stores instead of information consolidation in a data lake.” Gartner

12



http://www.gartner.com/newsroom/id/2809117


The Semantic Web and Schema.orgWhat does it mean for Enterprises?● 30% of all HTML pages contain structured data

(Source: https://commoncrawl.org/, 2014) ● 5 million web sites provide Schema.org data (Source: Google, 2014)

13

Sour

ce: T

FF20

15, C

hris

tian

Bize

r, U

ni

Man

nhei

m



https://commoncrawl.org/

http://www.slideshare.net/TourismFastForward/bizer-christian-schema-orgtourism





Search, analytics and new insightsThe IKEA principle

14

Mor

e: h

ttp://

ww

w.c

onflu

entfo

rms.

com

/201

5/05

/sem

antic

-sea

rch-

stra

tegy

.htm

l



http://www.confluentforms.com/2015/05/semantic-search-strategy.html


Semantic Web and Linked Data

15

+ + =Structured and

Unstructured Data Taxonomies Linked (Open) Data Semantic Search & Analytics




● ‘Interpretation’ of information demand by any means whatsoever (see: ‘semantic re-engineering’)

● Single information source● No explicit metadata, no standards● Application: Query expansion, Search refiners● Example: www.reegle.info/clean-energy-search

1-star Semantic Search

16



http://www.reegle.info/clean-energy-search


● ‘Interpretation’ of information demand based on a multi-lingual, controlled vocabulary including multi-faceted hierarchies

● Single information source● Explicit metadata layer● Application: Faceted search● Example: www.eip-water.eu/my-market-place/news


17



http://www.eip-water.eu/my-market-place/news


● ‘Interpretation’ of information demand based on multi-lingual, controlled vocabularies including multi-faceted hierarchies

● Multiple information sources incl. user profiles● Explicit and harmonised metadata layer, SKOS● Typical application: Integrated search /

personalisation● Example: www.healthdirect.gov.au


18



http://www.healthdirect.gov.au


● ‘Interpretation’ of information demand based on multi-lingual, interlinked KOS

● Multiple information sources transformed and stored in a Linked Data RDF Warehouse

● Explicit and harmonised metadata layer accessible via SPARQL endpoint

● Typical application: Business Analytics● Example: integrator.poolparty.biz/report_medicine


19



http://integrator.poolparty.biz/report_medicine/


● ‘Interpretation’ of information demand based on multi-lingual, interlinked KOS

● Multiple information sources transformed to RDF in real-time

● Explicit & harmonised metadata layer accessible via federated SPARQL endpoint


20

Bootstrapping a Virtual Semantic Data Warehouse




A Five-Star Rating System for Semantic Search

Precise interpretation of information demand

Multilingual, multi-faceted search

Multiple sources

SPARQL-enabled

Federated

21




The Rating System from a UX perspective

Higher recall

Higher precision

Less browsing, personalisation

New insights

Real-time

22




SPARQL is the new King of all Data Scientist’s tools

Read more on LinkedIn Pulse

23



https://www.linkedin.com/pulse/sparql-new-king-all-data-scientists-tools-andreas-blumauer

https://www.linkedin.com/pulse/sparql-new-king-all-data-scientists-tools-andreas-blumauer


Example: Disease Prevalence Analysis

The semantic approach for tracking scientific

publications

24



http://www.semantics.cc/aleksandar-kapisoda






http://integrator.poolparty.biz/report_medicine

http://integrator.poolparty.biz/report_medicine


Connect

Andreas Blumauer, MScITCEO, Semantic Web Company

[email protected] http://at.linkedin.com/in/andreasblumauer https://twitter.com/semwebcompany https://ablvienna.wordpress.com/

Sorry, no Facebook!

25



mailto:[email protected]

http://at.linkedin.com/in/andreasblumauer

https://twitter.com/semwebcompany

https://ablvienna.wordpress.com/

111 Blumauer - A Five-Star Rating System for Semantic Search

Documents