Mind the Web!

1
Mind the Web!

• Valentin Zacharias, Andreas Abecker, Imen Borgi,
  Simone Braun, Andreas Schmidt
• FZI Karlsruhe, Germany
• Denny Vrandecic
• AIFB, Universität Karlsruhe (TH), Germany
• Workshop on new forms of reasoning for the
  Semantic Web scalable, tolerant and dynamic
• Busan, Korea
• November 11, 2007

2
Thesis

• Not taking the web in Semantic Web serious has
  lead many Semantic Web researchers to do Semantic
  Systems and reasoning research without trying to
  tackle the fundamental challenges of the Semantic
  Web.

3
Overview

• Hypotheses
• Current trends and misconceptions
• Problems
• Some solutions

4
Hypotheses

1. Web scale is not just a bit larger
2. Ontologies are always changing
3. There is no right ontology

5
A triple
6
(No Transcript)
7
(No Transcript)
8
(No Transcript)
9
(No Transcript)
10
OWLIM 107 Triples 29
Suez Canal
11
RDF Store subsecond querying 108 Triples 25,26
Moon
Fensel / Harmelen estimate 1014 Triples
12
109 Triples
Earth
Fensel / Harmelen estimate 1014 Triples
13
1010 Triples
Jupiter
Fensel / Harmelen estimate 1014 Triples
14
1011 Triples
Fensel / Harmelen estimate 1014 Triples
15
1014 Triples
Distance Sun Pluto
Fensel / Harmelen estimate 1014 Triples
16
Hypotheses

1. Web scale is not just a bit larger
2. Ontologies are always changing
3. There is no right ontology

17
Dynamics of Web 2.0

• 100 edits in Wikipedia
• 200 tags in del.icio.us
• 270 image uploads to flickr
• 1100 blog entries
• per minute!
• No reason to believe the Semantic Web will be any
  less dynamic rather more!

18
Reasons for change

• Error correction
• New information
• Change of view
• Change of the world
• None of these reasons will disappear

19
Hypotheses

1. Web scale is not just a bit larger
2. Ontologies are always changing
3. There is no right ontology

20
There is no right ontology

• Many different views
• Conflicting views and interests
• Semantic desktop -gt many ontologies
• Ontologies are abstractions
• Different tasks lead to different ontologies
• Many design choices
• Google Base 100.000 schemas
• Intentionally false (Spam)

21
Misconceptions

1. Classical reasoning approaches will be used on
   the web
2. Assumption of correctness
3. It is a logical web, not a statistical one

22
Making weaker languages

• Reducing the expressive power of a logic does
  not solve any problems faster its only effect is
  to make some problems impossible to state.
• John Sowa

23
Classical Reasoning Approaches

• Even Polynomial is much to slow, unless a subset
  of statements is retrieved first
• Identifying this subset can be non trivial and is
  at the core of the challenge posed by the
  Semantic Web

24
Classical Reasoning Approaches Focus
theoretical worst cases

• Undecidability may be harmless in all cases
  that matter, completeness unachievable in
  practice anyway
• Worst Case complexity may be a bad guidance for
  usefulness.

Harmelen 2006
25
Assumption of correctness

• Reasoning is always on perfect statements
• Dealing with imperfection is preprocessing
• Vagueness
• Uncertainty
• Intentionally false
• Errors
• Outdated

26
A logical web

• Repeating a statement does not make it more true

27
(No Transcript)
28
(No Transcript)
29
Web Data Aggregation

• Statistical aggregation necessary for opinions
• Important in Information Retrieval (see PageRank)

30
A logical web

• Repeating a statement does not make it more true
• But more likely
• Rules in news three sources for a story
• Counting and relations matter
• Combine this with reasoning

31
Solutions?
32
See paper for more

• More approaches to reasoning
• KB partitioning and summarizing
• Approximate reasoning
• Massive parallelization
• Special purpose reasoners
• Deductive database techniques
• Combine retrieval and reasoning
• Emergence of Semantics
• Ontology Maturing
• Change on the Semantic Web

33
Conclusions

• Many promising approaches
• But which one to choose?
• Use cases are needed most!
• To evaluate the approaches
• To point into the right direction
• This is why this workshop is here

34
Thank you!
Not taking the web in Semantic Web serious has
lead many Semantic Web researchers to do Semantic
Systems and reasoning research without trying to
tackle the fundamental challenges of the Semantic
Web.