Towards Semantic Web Mining

1
TowardsSemanticWeb Mining
Bettina Berendt Institute of Information
Systems Humboldt University Berlin Andreas
Hotho, Gerd Stumme Institute AIFB University of
Karlsruhe
2
Overview

• 1 The Semantic Web and Web Mining
• Web Mining
• Semantic Web
• 2 Extracting Semantics from the Web
• 3 Mining the Semantic Web
• 4 Closing the Loop

This presentation shows the general vision
whereas the paper in the proceedings focusses on
usage mining.
3
What is Web Mining?

• Goals
• the improvement of site design and site
  structure,
• the generation of dynamic recommendations,
• and improving marketing.

• Web Mining Areas
• Web content mining

4
Limitations of Web Mining ...

• Structure/Content Mining
• Distributed content is not related
• Content encoded in structure is not used
• Content only understandable for humans
• Huge amount of data

• Usage Mining
• Usage tracking by technical access logs (urls)
• mapping of urls to general concepts is
  difficult and expensive
• events and urls are n-to-m

5
What is the Semantic Web?

• The current WWW is a great success with respect
  to
• the amount of available information
• the number of users

• However, one problem of the current WWW is that
  the information may only be interpreted by
  humans.
• The Semantic Web tries to overcome this problem
  by using machine-processable metadata and
  ontologies.

• Goals include
• more effective retrieval of Web information
• knowledge-based inference

6

• cooperateswith(X,Y)
• cooperateswith(Y,X)

TOP
NAME
PERSON
PERSON
TITEL
PROJECT
COOPERATES
--
COOPERATES
--
WITH
WITH
Ontology
WORKS-IN
RESEARCHER
RESEARCHER
Semantic Web Mining
Andreas Hotho
WORKS-IN
DAMLPROJ
URI-SWMining
-
Relational Metadata
URI-AHO
WORKS-IN
COOPERATES
-
COOPERATES
-
WITH
WITH
URI-GST
WWW
7
The Semantic Web and Web Mining
What are the benefits of bringing the Semantic
Web and Web Mining together?
WWW

• The results of Web Mining can be improved by
  exploiting the new semantical structures in the
  Web.
• Web Mining can help to build the Semantic Web.

Web Mining
Semantic Web
8
Overview

• 1 The Semantic Web and Web Mining
• Web Mining
• Semantic Web
• 2 Extracting Semantics from the Web
• Ontology Learning
• Instance Learning
• 3 Mining the Semantic Web
• 4 Closing the Loop

9
Extracting Semantics from the Web

• Web Mining can help
• to learn structures for knowledge organization
  (e.g. ontologies)
• and to populate them.

Ontology Learning
Instance Learning

• The approaches are semi-automatic
• A lot of tacit background knowledge, experiences,
  social conventions, etc is involved in the
  modeling process.
• In order to obtain high quality results, a human
  has to be in the loop.
• If this were not the case, the Semantic Web would
  be superfluous!

10
Example
Ontology Learning
is-a hierarchy
Mädche, Staab ECAI 2000
11
Instance Learning

• Providers are not willing to annotate too many
  pages manually.
• Moreover users may need to extract more/other
  information as provided.

• Information Extraction is a set of
  (semi-)automatice methods for locating important
  facts in electronic documents.
• IE may be used for instance to support human
  annotators in locating relevant facts in
  documents via information highlighting. Tools are
  e.g.
• FASTUS Hobbs et al 1996
• OntoMat Annotizer Handschuh, Staab 2002.

12
Example
Information Highlighting for supporting
annotation based on IE techniques.
13
Example
Crawling the semantic web for filling the ontology
ontology
ontology
score
metadata
count
rating
2. relevance
1. validation
relational, sum
ltfClass rdfabout"cairbus"gt
ltcairline rdfabout"lufthansa"gt
ltcowns
ontology
ontology
rdfresource"airbus123"/gt
3. summarization
lt/cairlinegt
text
count
rating
Lufthansa just received its
newest Airbus A340 from the
base in Toulouse. Airbus
Industries added some new
1. lexicon lookup
2. relevance
relational, sum
features to this version of the
airplane.
Ehrig et al, 2002
14
Overview

• 1 The Semantic Web and Web Mining
• Web Mining
• Semantic Web
• 2 Extracting Semantics from the Web
• Ontology Learning
• Instance Learning
• 3 Mining the Semantic Web
• Semantic Web Content/Structure Mining
• Semantic Web Usage Mining
• 4 Closing the Loop

15
Exploiting Semantics for Web Content Mining
Example
Conceptual Clustering of Emails (and Bookmarks)
using IE and Formal Concept Analysis for
supporting navigation and retrieval.
16
Example
Semantic Web Structure/Content Mining
Knowledge base Hotel Wellnesshotel GolfCourse
Seaview belongsTo(Seaview, Wellnesshotel) ...
17
Semantic Web Usage Mining
Basic idea associate each requested page with
one or more ontological entities,
to better understand the process of
navigation
From logfile analysis ...

• Use the gained knowledge to
• understand search strategies
• improve navigation design
• personalization

p3ee24304.dip.t-dialin.net - - 19/Mar/20021203
51 0100 "GET /search.html?lostsee20stran
dsyn023785ordasc HTTP/1.0" 200 1759
p3ee24304.dip.t-dialin.net - -
19/Mar/2002120506 0100 "GET
/search.html?lostsee20strandplowsyn023785or
ddesc HTTP/1.0" 200 8450 p3ee24304.dip.t-dialin.n
et - - 19/Mar/2002120641 0100 "GET
/mlesen.html?Item3456syn023785 HTTP/1.0" 200
3478
... to semantic logfile analysis
Refine search
Choose item
Search by Location and Price
Search by Location
Look at individual Hotel.
Berendt Spiliopoulou 2000 Berendt 2002
Oberle 2001
18
Overview

• 1 The Semantic Web and Web Mining
• Web Mining
• Semantic Web
• 2 Extracting Semantics from the Web
• Ontology Learning
• Instance Learning
• 3 Mining the Semantic Web
• Semantic Web Content/Structure Mining
• Semantic Web Usage Mining
• 4 Closing the Loop

19
Example
Mining the web for learning ontologies, ...
is-a hierarchy
Mädche, Staab ECAI 2000
20
Example
..., mining the web for filling the ontology, ...
21
Example
... and using the ontology for mining again.
Knowledge base Hotel Wellnesshotel GolfCourse
Seaview belongsTo(Seaview, Wellnesshotel) ...
22
Overview

• 1 The Semantic Web and Web Mining
• 2 Extracting Semantics from the Web
• Mining the Semantic Web
• Closing the Loop

The End

• Conclusion
• Towards Semantic Web Mining
• The results of Web Mining can be improved by
  exploiting the new semantical structures in the
  Web.
• Web Mining can help to build the Semantic Web.
• Both aspects can be combined by closing the loop.