Event Notification for Semantic Web Documents using Distributed Authoring

1
Event Notification for Semantic Web Documents
using Distributed Authoring
Reuben Evans Annika Hinze University of
Waikato Presentation at the 1st NZ Workshop on
Ontologies and the Semantic web 4 March 2006
2
Motivation Event Notification
Notify about items of interest XML books by
Smith
pub(item,...)
Filtering
3
Motivation Change in the Semantic Web

• Typical interaction Querying the Semantic Web
• How do we know if data changed after querying?
• Register continuous query at an ENS
• Notification on event

2. Agent consults information sources
1. User makes request
3. Agent compiles answer
4
Motivation Change in the Semantic Web

• User/Agent interestdefined in profiles
• Profiles continuous queries, interest in
  changes over time
• Types of events
• New page
• Changed data
• Changed metadata
• Deleted data

5
Contents

• Motivation
• Design of ENS for Semantic Web
• Evaluation
• Conclusion Further work

6
Design Representing RDF

• As a Graph

• As N-TriplesLecturer Teaches Course.
• COMP582 a Course.
• Annika Teaches COMP582.
• Annika a Lecturer.
• Teaches a rdfProperty.

As XML ltrdfRDF xmlnsrdf
xmlnsrdfsgt ltrdfProperty rdfaboutTeaches
/gt ltCourse rdfaboutCOMP582/gt ltLecturer
rdfaboutAnnikagt ltTeaches rdfrescouceCOMP
582/gt lt/Lecturergt lt/rdfRDFgt
7
Design Decisions

• ENS well known concept
• Focus on attribute value pairs
• Sometimes on XML
• Requires event messages sent in
• Observation of change missing
• Distributed Authoring Systems (DAS)
• Mechanism for version control
• For plan text
• Structuring missing

8
Design Architecture
Document Author
Document User
RDF Changes
Profiles
Notifications
Repository (SVN)
Delta Adapter
Profile Filter
ENS for Semantic Web
9
Repository

• Subversion
• CVS relative
• Versioning software
• Uses text diff to store and compare documents
• Here used as trigger and versioning of RDF/XML
  files

Original Document ltrdfRDF xmlnsrdf
xmlnsrdfsgt ltrdfProperty rdfaboutTeaches
/gt ltCourse rdfaboutCOMP582/gt ltLecturer
rdfaboutAnnikagt ltTeaches rdfresourceCOMP
582/gt lt/Lecturergt lt/rdfRDFgt
New Document ltrdfRDF xmlnsrdf
xmlnsrdfsgt ltrdfProperty rdfaboutTeaches
/gt ltCourse rdfaboutCOMP582/gt ltLecturer
rdfaboutAnnikagt ltTeaches rdfresourceCOMP
319/gt lt/Lecturergt lt/rdfRDFgt
10
Delta-Adapter

• Takes copies of documents
• Performs XML diff using Delta XML
• XSLT transform on result to convert back to RDF
• Adds action triples describing changes
  (annotations)
• Cuts out unchanged triples
• Sends results to filter

As XML ltrdfRDF xmlnsdeltaxml" xmlnsrdf"
deltaxmldelta"WFmodify"gt ltrdfProperty
deltaxmldelta"unchanged" rdfabout"Teaches"/gt
ltCourse deltaxmldelta"unchanged"
rdfabout"COMP582"/gt ltLecturer
deltaxmldelta"WFmodify" rdfabout"Annika"gt lt
Teaches deltaxmldelta"WFmodify
deltaxmlold-attributes"rdfrescoucequotCOM
P582quot" deltaxmlnew-attributes"rdfresour
cequotCOMP319quot"/gt lt/Lecturergt lt/rdfRDFgt
11
Profile Filter

• Based on ApproXFilter
• Tree-based matching
• Can talk to any number of Delta Adapters
• Here used on changed document sections with
  annotations

Filter Input ( Delta Output) ltAnnikagt
ltactiongtmodifylt/actiongt ltTeachesgt COMP582
ltactiongtdeletelt/actiongt lt/Teachesgt
ltTeachesgt COMP319 ltactiongtaddlt/actiongt
lt/Teachesgt lt/Annikagt
12
Filter Language

• Three level language
• Matches triple structure of RDF
• Allows conjuncts at any level
• Each profile has an accuracy value

Filter Input ltAnnikagt ltactiongtmodifylt/action
gt ltTeachesgt COMP582 ltactiongtdeletelt/action
gt lt/Teachesgt ltTeachesgt COMP319
ltactiongtaddlt/actiongt lt/Teachesgt lt/Annikagt
Example Annika Teaches AND Teaches
actionadd
13
Contents

• Motivation
• Design of ENS for Semantic Web
• Evaluation
• Conclusion Further work

14
Evaluation Observer

• Observation time
• O( document size)
• One triple removed
• All triples removed
• (filtering over initialization after 20,000
  triples)
• O(Syntactic changes)

15
Evaluation Filter

• Event parsing time
• O(distinct profiles)
• No affect of number of conjunctions in profiles
• Event matching time
• O(distinct profiles)
• Linear increase if conjunct at end of profile
• No affect if conjunct joins two distinct triples
  (due to cost calculation)

16
Contents

• Motivation
• Design of ENS for Semantic Web
• Evaluation
• Conclusion Further work

17
Conclusion

• Designed and Implemented an ENS for detecting
  changes in RDF/XML files
• Exploits the XML structure XML diff and filter
• Supports change/delete/new
• Did not talk about
• Approximative filtering using synonym list
• Cost measurements
• User Interface to input profiles as tree

18
Further Work

• Work on the problem of distribution
• Can currently only talk to one filter server
• Distribute profile filtering, allow profiles to
  span RDF from multiple repositories
• Add support for XML namespaces
• Extensive comparison to triple and graph-based
  approaches

19
Thank you for your attention!

• Questions?
• Why not work on the graph?
• Measure of robustness

20
References1

• 1 Dan Connolly, Frank van Harmelen, Ian
  Horrocks, Deborah L. McGuinness and Peter F.
  Patel-Schneider, DAMLOIL (March 2001) Reference
  Description, W3C Note 18 December 2001, visited
  on March 24, 2005, http//www.w3.org/TR/2001/NOTE-
  damloil-reference-20011218
• 2 L. Qin and V. Atluri, Ontology-guided
  Change Detection to the Semantic Web Data,'' 23rd
  International Conference on Conceptual Modeling
  (ER2004), November, 2004.
• 3 Peter F. Patel-Schneider, Patrick Hayes, and
  Ian Horrocks OWL Web Ontology Language Semantics
  and Abstract Syntax, W3C Recommendation 10
  February 2004, visited on March 24, 2005,
  http//www.w3.org/TR/2004/REC-owl-semantics-200402
  10/
• 4 Jeff Heflin OWL Web Ontology Language Use
  Cases and Requirements, W3C Recommendation 10
  February 2004, visited on March 24, 2005,
  http//www.w3.org/TR/2004/REC-webont-req-20040210/
  
• 5Graham Klyne and Jeremy J. Carroll, Resource
  Description Framework (RDF) Concepts and
  Abstract Syntax, W3C Recommendation, 10 February
  2004, http//www.w3.org/TR/2004/REC-rdf-concepts-2
  0040210/

21
References2

• 6 Dan Brickley and R.V. Guha, RDF Vocabulary
  Description Language 1.0 RDF Schema, W3C
  Recommendation 10 February 2004 visited on March
  23, 2005, http//www.w3.org/TR/2004/REC-rdf-schema
  -20040210/
• 7 Dave Beckett, RDF/XML Syntax Specification,
  W3C Recommendation, 10 February 2004, visited on
  March 23, 2005, http//www.w3.org/TR/2004/REC-rdf-
  syntax-grammar-20040210/
• 8Tim Berners-Lee, James Hendler, Ora Lassila
  The Semantic Web Scientific American May 2001
• 9 Yann Michel ApproxFilter an approximative,
  XML-based event filter, Student research project
  Free University Berlin 2005
• 10 Robin La Fontaine "A Delta Format for XML
  Identifying Changes in XML Files and Representing
  the Changes in XML" XML Europe, Berlin, May 2001
• 11Ben Collins-Sussman, Brian W. Fitzpatrick, C.
  Michael Pilato, Version Control with Subversion
  ORielly  June 2004 ISBN 0-596-00448-6

22
Motivation Approximative Filtering

• Problem too rigid subscriptions
• Keyword-based systems XML and Smith
• What about book/article/journal and
• editor/author/reference smith or
• smith as reference/acknowledgement/editor
• XML-based systems (e.g. using XPath)
• What about the other options?
• What if structure not known?
• How to serve several sources?

/catalog/booktitle XML" and author Smith"
23
Motivation Approximative Filtering

• Solution Approximative Filtering
• Subscription using ApproxQL language
• Profile query
• Profile transformations

book title "XML" and author "Smith"
24
Road map

• Motivation
• ApproxFilter Algorithms
• Evaluation
• Summary and Future Work

25
Structure of the algorithms

• Step 1 Normalization (conjunctive normal form)
• Step 2 - Profile Extension (renaming, skipping,
  insertion)
• Step 3 Subscription Tree-building
• Step 4 Filtering document against DAG
• depth-first
• Cost accumulation

26
ApproxFilter Algorithms

• Principle
• Sequentially following the document follow
  paths in the DAG
• Cost calculation while following the DAG upwards
• Name/Value insert costs into linked profile
  nodes
• Several links/nodes to foolow choose minimal
  costs
• End-Tag Combine using Boolean operators
• Invalidate Costs of child nodes
• Two algorithms
• Cost optimized
• Space optimized

27
ApproxFilter Algorithms DAG
profiles booktitle"XML" and
authorsmith" booktitle"DB" and "XML" and
year2000
DB
abstract
RDF
title
match DAG
XML
article


book
author
Fred
2000
year
renamings book ? article,
title ? abstract, XML ?
RDF deletions title, XML insertions as
neccesary
profile trees
title
XML
book
author
Fred
DB
title
XML
book
year
2000
28
Data structures time optimized
abstract
DB
author
RDF
article
DB
XML
book
RDF

title
Fred
XML
2000
year
Blue links point to hashes with all keys allowed
at these places Red links have renaming/deletion
costs attached Example red link of key "RDF"
has deletion costs of "title" as well
as renaming costs from "XML" to "RDF"
title
XML
book
author
Fred
DB
title
XML
book
year
2000
Space O(n2)
29
Data structures time optimized
abstract
DB
author
RDF
article
title
XML
book
year


Fred
2000
Green keys show deletions and may have to be
followed transitively if no hash key is found.
Several green links are possible. Links have
renaming/deletion costs attached Example green
link has deletion costs of "title". If also
author could be deleted Another green link
would go the the Hash with the key "Fred".
title
XML
book
author
Fred
DB
title
XML
book
year
2000
Space O(n)
30
Time-optimized ApproxFilter
abstract
DB
author
RDF
article
DB
XML
book
RDF

title
Fred
ltdocgt ltarticlegt ltabstractgt RDF ... XML
lt/abstractgt ltauthorgtFredlt/authorgt
ltyeargt2000lt/yeargt lttitlegt ... RDF ... DB
lt/titlegt lt/articlegt ltbookgt
ltyeargt2000lt/yeargt lttitlegt RDF ... DB ...
lt/titlegt ltcommentgt ... XML lt/commentgt
lt/bookgt lt/docgt
XML
2000
year
title
XML
book
author
Fred
DB
title
XML
book
year
2000
31
Time-optimized ApproxFilter
abstract
DB
author
RDF
article
DB
XML
book
RDF

title
Fred
ltdocgt ltarticlegt ltabstractgt RDF ... XML
lt/abstractgt ltauthorgtFredlt/authorgt
ltyeargt2000lt/yeargt lttitlegt ... RDF ... DB
lt/titlegt lt/articlegt ltbookgt
ltyeargt2000lt/yeargt lttitlegt RDF ... DB ...
lt/titlegt ltcommentgt ... XML lt/commentgt
lt/bookgt lt/docgt
XML
2000
year
title
XML
book
author
Fred
DB
title
XML
book
year
2000
32
Time-optimized ApproxFilter
abstract
DB
author
RDF
article
DB
XML
book
RDF

title
Fred
ltdocgt ltarticlegt ltabstractgt RDF ... XML
lt/abstractgt ltauthorgtFredlt/authorgt
ltyeargt2000lt/yeargt lttitlegt ... RDF ... DB
lt/titlegt lt/articlegt ltbookgt
ltyeargt2000lt/yeargt lttitlegt RDF ... DB ...
lt/titlegt ltcommentgt ... XML lt/commentgt
lt/bookgt lt/docgt
XML
2000
year
title
XML
book
author
Fred
DB
title
XML
book
year
2000
33
Time-optimized ApproxFilter
abstract
DB
author
RDF
article
DB
XML
book
RDF

title
Fred
ltdocgt ltarticlegt ltabstractgt RDF ... XML
lt/abstractgt ltauthorgtFredlt/authorgt
ltyeargt2000lt/yeargt lttitlegt ... RDF ... DB
lt/titlegt lt/articlegt ltbookgt
ltyeargt2000lt/yeargt lttitlegt RDF ... DB ...
lt/titlegt ltcommentgt ... XML lt/commentgt
lt/bookgt lt/docgt
XML
2000
year
title
XML
book
author
Fred
DB
title
XML
book
year
2000
34
Time-optimized ApproxFilter
abstract
DB
author
RDF
article
DB
XML
book
RDF

title
Fred
ltdocgt ltarticlegt ltabstractgt RDF ... XML
lt/abstractgt ltauthorgtFredlt/authorgt
ltyeargt2000lt/yeargt lttitlegt ... RDF ... DB
lt/titlegt lt/articlegt ltbookgt
ltyeargt2000lt/yeargt lttitlegt RDF ... DB ...
lt/titlegt ltcommentgt ... XML lt/commentgt
lt/bookgt lt/docgt
XML
2000
year
title
XML
book
(cost 7)
author
Fred
DB
title
XML
book
year
2000
renaming XML -gt RDF
35
Time-optimized ApproxFilter
abstract
DB
author
RDF
article
DB
XML
book
RDF

title
Fred
ltdocgt ltarticlegt ltabstractgt RDF ... XML
lt/abstractgt ltauthorgtFredlt/authorgt
ltyeargt2000lt/yeargt lttitlegt ... RDF ... DB
lt/titlegt lt/articlegt ltbookgt
ltyeargt2000lt/yeargt lttitlegt RDF ... DB ...
lt/titlegt ltcommentgt ... XML lt/commentgt
lt/bookgt lt/docgt
XML
2000
year
title
XML
book
(cost 0 min(7,0))
author
Fred
DB
title
XML
book
year
2000
36
Time-optimized ApproxFilter
abstract
DB
author
RDF
article
DB
XML
book
RDF

title
Fred
ltdocgt ltarticlegt ltabstractgt RDF ... XML
lt/abstractgt ltauthorgtFredlt/authorgt
ltyeargt2000lt/yeargt lttitlegt ... RDF ... DB
lt/titlegt lt/articlegt ltbookgt
ltyeargt2000lt/yeargt lttitlegt RDF ... DB ...
lt/titlegt ltcommentgt ... XML lt/commentgt
lt/bookgt lt/docgt
XML
2000
year
title
XML
book
(reset to cost 0)
(cost 4 4 0 0)
author
Fred
DB
title
XML
book
year
2000
renaming title -gt abstract
child costs
insert (skip) costs
37
Time-optimized ApproxFilter
abstract
DB
author
RDF
article
DB
XML
book
RDF

title
Fred
ltdocgt ltarticlegt ltabstractgt RDF ... XML
lt/abstractgt ltauthorgtFredlt/authorgt
ltyeargt2000lt/yeargt lttitlegt ... RDF ... DB
lt/titlegt lt/articlegt ltbookgt
ltyeargt2000lt/yeargt lttitlegt RDF ... DB ...
lt/titlegt ltcommentgt ... XML lt/commentgt
lt/bookgt lt/docgt
XML
2000
year
title
XML
book
(cost 4)
author
Fred
DB
title
XML
book
year
2000
38
Time-optimized ApproxFilter
abstract
DB
author
RDF
article
DB
XML
book
RDF

title
Fred
ltdocgt ltarticlegt ltabstractgt RDF ... XML
lt/abstractgt ltauthorgtFredlt/authorgt
ltyeargt2000lt/yeargt lttitlegt ... RDF ... DB
lt/titlegt lt/articlegt ltbookgt
ltyeargt2000lt/yeargt lttitlegt RDF ... DB ...
lt/titlegt ltcommentgt ... XML lt/commentgt
lt/bookgt lt/docgt
XML
2000
year
title
XML
book
(cost 4)
author
Fred
(cost 0)
DB
title
XML
book
year
2000
39
Time-optimized ApproxFilter
abstract
DB
author
RDF
article
DB
XML
book
RDF

title
Fred
ltdocgt ltarticlegt ltabstractgt RDF ... XML
lt/abstractgt ltauthorgtFredlt/authorgt
ltyeargt2000lt/yeargt lttitlegt ... RDF ... DB
lt/titlegt lt/articlegt ltbookgt
ltyeargt2000lt/yeargt lttitlegt RDF ... DB ...
lt/titlegt ltcommentgt ... XML lt/commentgt
lt/bookgt lt/docgt
XML
2000
year
title
XML
book
(cost 4)
author
Fred
(reset to cost 0)
(cost 0)
DB
title
XML
book
year
2000
40
Time-optimized ApproxFilter
abstract
DB
author
RDF
article
DB
XML
book
RDF

title
Fred
ltdocgt ltarticlegt ltabstractgt RDF ... XML
lt/abstractgt ltauthorgtFredlt/authorgt
ltyeargt2000lt/yeargt lttitlegt ... RDF ... DB
lt/titlegt lt/articlegt ltbookgt
ltyeargt2000lt/yeargt lttitlegt RDF ... DB ...
lt/titlegt ltcommentgt ... XML lt/commentgt
lt/bookgt lt/docgt
XML
2000
year
title
XML
book
(cost 4)
author
Fred
(cost 0)
DB
title
XML
book
year
2000
41
Time-optimized ApproxFilter
abstract
DB
author
RDF
article
DB
XML
book
RDF

title
Fred
ltdocgt ltarticlegt ltabstractgt RDF ... XML
lt/abstractgt ltauthorgtFredlt/authorgt
ltyeargt2000lt/yeargt lttitlegt ... RDF ... DB
lt/titlegt lt/articlegt ltbookgt
ltyeargt2000lt/yeargt lttitlegt RDF ... DB ...
lt/titlegt ltcommentgt ... XML lt/commentgt
lt/bookgt lt/docgt
XML
2000
year
title
XML
book
(cost 4)
author
Fred
(cost 0)
DB
title
XML
book
year
2000
42
Time-optimized ApproxFilter
abstract
DB
author
RDF
article
DB
XML
book
RDF

title
Fred
ltdocgt ltarticlegt ltabstractgt RDF ... XML
lt/abstractgt ltauthorgtFredlt/authorgt
ltyeargt2000lt/yeargt lttitlegt ... RDF ... DB
lt/titlegt lt/articlegt ltbookgt
ltyeargt2000lt/yeargt lttitlegt RDF ... DB ...
lt/titlegt ltcommentgt ... XML lt/commentgt
lt/bookgt lt/docgt
XML
2000
year
title
XML
book
(cost 4)
author
Fred
(cost 0)
DB
title
XML
book
year
2000
(cost 0)
43
Time-optimized ApproxFilter
abstract
DB
author
RDF
article
DB
XML
book
RDF

title
Fred
ltdocgt ltarticlegt ltabstractgt RDF ... XML
lt/abstractgt ltauthorgtFredlt/authorgt
ltyeargt2000lt/yeargt lttitlegt ... RDF ... DB
lt/titlegt lt/articlegt ltbookgt
ltyeargt2000lt/yeargt lttitlegt RDF ... DB ...
lt/titlegt ltcommentgt ... XML lt/commentgt
lt/bookgt lt/docgt
XML
2000
year
title
XML
book
(cost 4)
author
Fred
(cost 0)
DB
title
XML
book
year
2000
(reset to cost 0)
(cost 0)
44
Time-optimized ApproxFilter
abstract
DB
author
RDF
article
DB
XML
book
RDF

title
Fred
ltdocgt ltarticlegt ltabstractgt RDF ... XML
lt/abstractgt ltauthorgtFredlt/authorgt
ltyeargt2000lt/yeargt lttitlegt ... RDF ... DB
lt/titlegt lt/articlegt ltbookgt
ltyeargt2000lt/yeargt lttitlegt RDF ... DB ...
lt/titlegt ltcommentgt ... XML lt/commentgt
lt/bookgt lt/docgt
XML
2000
year
title
XML
book
(cost 4)
author
Fred
(cost 0)
DB
title
XML
book
year
2000
45
Time-optimized ApproxFilter
abstract
DB
author
RDF
article
DB
XML
book
RDF

title
Fred
ltdocgt ltarticlegt ltabstractgt RDF ... XML
lt/abstractgt ltauthorgtFredlt/authorgt
ltyeargt2000lt/yeargt lttitlegt ... RDF ... DB
lt/titlegt lt/articlegt ltbookgt
ltyeargt2000lt/yeargt lttitlegt RDF ... DB ...
lt/titlegt ltcommentgt ... XML lt/commentgt
lt/bookgt lt/docgt
XML
2000
year
title
XML
book
(cost 4)
author
Fred
(cost 0)
DB
title
XML
book
year
2000
46
Time-optimized ApproxFilter
abstract
DB
author
RDF
article
DB
XML
book
RDF

title
Fred
ltdocgt ltarticlegt ltabstractgt RDF ... XML
lt/abstractgt ltauthorgtFredlt/authorgt
ltyeargt2000lt/yeargt lttitlegt ... RDF ... DB
lt/titlegt lt/articlegt ltbookgt
ltyeargt2000lt/yeargt lttitlegt RDF ... DB ...
lt/titlegt ltcommentgt ... XML lt/commentgt
lt/bookgt lt/docgt
XML
2000
year
title
XML
book
(cost 4)
(cost 7)
author
Fred
(cost 0)
DB
title
XML
book
year
2000
47
Time-optimized ApproxFilter
abstract
DB
author
RDF
article
DB
XML
book
RDF

title
Fred
ltdocgt ltarticlegt ltabstractgt RDF ... XML
lt/abstractgt ltauthorgtFredlt/authorgt
ltyeargt2000lt/yeargt lttitlegt ... RDF ... DB
lt/titlegt lt/articlegt ltbookgt
ltyeargt2000lt/yeargt lttitlegt RDF ... DB ...
lt/titlegt ltcommentgt ... XML lt/commentgt
lt/bookgt lt/docgt
XML
2000
year
title
XML
book
(cost 4)
(cost 7)
author
Fred
(cost 0)
DB
title
XML
book
year
2000
48
Time-optimized ApproxFilter
abstract
DB
author
RDF
article
DB
XML
book
RDF

title
Fred
ltdocgt ltarticlegt ltabstractgt RDF ... XML
lt/abstractgt ltauthorgtFredlt/authorgt
ltyeargt2000lt/yeargt lttitlegt ... RDF ... DB
lt/titlegt lt/articlegt ltbookgt
ltyeargt2000lt/yeargt lttitlegt RDF ... DB ...
lt/titlegt ltcommentgt ... XML lt/commentgt
lt/bookgt lt/docgt
XML
2000
year
title
XML
book
(cost 4 min(4,070))
(reset to cost 7)
author
Fred
(cost 0)
DB
title
XML
book
year
2000
49
Time-optimized ApproxFilter
abstract
DB
author
RDF
article
DB
XML
book
RDF

title
Fred
ltdocgt ltarticlegt ltabstractgt RDF ... XML
lt/abstractgt ltauthorgtFredlt/authorgt
ltyeargt2000lt/yeargt lttitlegt ... RDF ... DB
lt/titlegt lt/articlegt ltbookgt
ltyeargt2000lt/yeargt lttitlegt RDF ... DB ...
lt/titlegt ltcommentgt ... XML lt/commentgt
lt/bookgt lt/docgt
XML
2000
year
title
XML
book
(cost 4)
author
Fred
(cost 9 5 4 0 0)
(cost 0)
DB
title
XML
book
year
2000
renaming book -gt article
child costs
insert (skip) costs
50
Road map

• Motivation
• ApproxFilter Algorithms
• Implementation
• Evaluation
• Summary and Future Work

51
ApproxFilter Implementation

• Profile service
• Document service
• Parallel execution

52
ApproxFilter Implementation

• Data structure as implemented
• Simplified version of space optimized algorithm
• compact profile tree (bottom)
• match structure for filtering document
  structures (left)
• content-synonym set (top)

53
Road map

• Motivation
• ApproxFilter Algorithms
• Implementation
• Evaluation
• Summary and Future Work

54
Functional Evaluation

• Matching with and without transformations
• Example data from amazon.com

55
Performance Evaluation

• Time for profile evaluation

56
Performance Evaluation

• Space for data structures
• O(n)

57
Road map

• Motivation
• ApproxFilter Algorithms
• Implementation
• Evaluation
• Summary and Future Work

58
Summary

• First approximative filtering for XML documents
• Approximative language for subscriptions in a
  publish/subscribe system
• Design of filter algorithm
• Two implementations
• proof of concept ApproXFilter prototype
• Initial qualitative and quantitative testing

59
Future Work

• Complex definition of cost, application dependent
• Use of user relevance feedback to adjust the
  costs
• use of domain ontologies and personalised
  ontologies.
• Extension of prototype to support disjunctions
• Further analysis of proposed algorithms
• ApproXFilter in the context of digital library
  software

60
Thank you for your attention! And nowquestions.

• Contact
• Annika Hinze
• hinze_at_cs.waikato.ac.nz

61
Grundprinzip
queries booktitle"XML" and author"Fred" (pr
ofiles) booktitle"DB" and "XML" and year2000
renamings book ? article, title ? abstract, XML
? RDF
DB
abstract
deletions title, XML insertions beliebig
RDF
title
match DAG
document ltdocgt ltarticlegt ltabstractgt RDF
... XML lt/abstractgt ltauthorgtFredlt/authorgt
ltyeargt2000lt/yeargt lttitlegt ... RDF ... DB
lt/titlegt lt/articlegt ltbookgt
ltyeargt2000lt/yeargt lttitlegt RDF ... DB ...
lt/titlegt ltcommentgt ... XML lt/commentgt
lt/bookgt lt/docgt
XML
article


book
author
Fred
year
2000
profile/query trees
title
XML
book
author
Fred
DB

• algorithmus (Prinzip)
• Abstieg in Doc folge Pfaden im Match-DAG
  Kostenberechung während Aufstieg
• Name/wert Trage Kosten in verlinkte
  Profile-Knoten ein
• Mehrere Links/Zielknoten wähle beste Kosten
• End-Tag (z.B. von title) Kombiniere anhand von
  Logischen Operatoren (in jedem Kandidat-Profil)
• Kosten-Invalidierung der Kinder

title
XML
book
year
2000
Links (rot) sind teilweise dargestelllt Zu jedem
Namen/Wert werden irgendwo Kosten festgehalten,
die über Link in jeweiligen Profile-Knoten
eingetragen werden.