10
Supporting Users in Finding Relevant Sources of Linked Open Data at Web-Scale Thomas Gottron , Ansgar Scherp, Bastian Krayer, Arne Peters Get the Feeling!

Get the Google Feeling! Supporting Users in Finding Relevant Sources

Embed Size (px)

DESCRIPTION

Presentation of the LODatio system in the round of finalists at the Semantic Web Challenge (Billion Triples Track) at ISWC 2012

Citation preview

Page 1: Get the Google Feeling! Supporting Users in Finding Relevant Sources

Supporting Users in Finding Relevant Sources of Linked Open Data at Web-Scale

Thomas Gottron, Ansgar Scherp, Bastian Krayer, Arne Peters

Get the Feeling!

Page 2: Get the Google Feeling! Supporting Users in Finding Relevant Sources

Thomas Gottron BTC 2012 2Get the Google Feeling

System Support for Searchers

Bates, M.J.: Where should the person stop and the information search interface start? Information Processing and Management 26(5), 575–591 (1990)

none

displayoptions

execute oncommand

automaticexecution

System involvement

Mov

e User activityTa

ctic

Strata

gem

Strate

gy

monitor andrecommend

OperationalSystems

(then)

Area of recommended development

Hold for later

Hold for later

(skip)

Pure user activity

Page 3: Get the Google Feeling! Supporting Users in Finding Relevant Sources

Thomas Gottron BTC 2012 3Get the Google Feeling

System Support Helps: Query Specific Snippets

Tombros, A., Sanderson, M.: Advantages of query biased summaries in information retrieval.SIGIR’98. pp. 2–10 (1998)

Recall Precision

Speed Satisfaction

Page 4: Get the Google Feeling! Supporting Users in Finding Relevant Sources

Thomas Gottron BTC 2012 4Get the Google Feeling

System Support Helps: Query Suggestions

Kelly, D., et al. Effects of popularity and quality on the usage of query suggestions during information search. CHI '10, p 45-54, (2010)

41% of all queries were chosen from suggestions

Find entry point Think out of the box Identify new query terms

Page 5: Get the Google Feeling! Supporting Users in Finding Relevant Sources

Thomas Gottron BTC 2012 5Get the Google Feeling

Page 6: Get the Google Feeling! Supporting Users in Finding Relevant Sources

Did you mean?

Result Set Size

Result Snippets

Related Queries

Ranked Retrieval

Page 7: Get the Google Feeling! Supporting Users in Finding Relevant Sources

Thomas Gottron BTC 2012 7Get the Google Feeling

Schema-based Index Design

Page 8: Get the Google Feeling! Supporting Users in Finding Relevant Sources

Thomas Gottron BTC 2012 8Get the Google Feeling

„Under the hood“

SPARQL

Snippets

Generalize

Retrieve Datasources

Query translation

Rank

Specify

Count

Select

Select

• 1 query for result set and result set size• N queries for ranking data and snippets• 2 queries per related query

Page 9: Get the Google Feeling! Supporting Users in Finding Relevant Sources

Thomas Gottron BTC 2012 9Get the Google Feeling

Stats

Use of the complete BTC 2012 dataset Index size

133M schema triples 224M payload triples

Commodity hardware Data processing LODation service provision

Index construction (15h) and optimization (5h) Response time: < 1s on a single CPU machine

Page 10: Get the Google Feeling! Supporting Users in Finding Relevant Sources

Get the Feeling!

Thank you!