ASSAM: A Tool for Semi-Automatically Annotating Semantic Web - - PowerPoint PPT Presentation

▶
assam a tool for semi automatically annotating semantic
SMART_READER_LITE
LIVE PREVIEW

ASSAM: A Tool for Semi-Automatically Annotating Semantic Web - - PowerPoint PPT Presentation

Jan 15, 2024 580 likes •1.27k views

Outline Introduction ASSAM Annotator OATS Conclusion ASSAM: A Tool for Semi-Automatically Annotating Semantic Web Services Andreas He, Eddie Johnston and Nicholas Kushmerick University College Dublin, Ireland November 10, 2004 4 ,

slide-1
SLIDE 1

Outline Introduction ASSAM Annotator OATS Conclusion

ASSAM: A Tool for Semi-Automatically Annotating Semantic Web Services

Andreas Heß, Eddie Johnston and Nicholas Kushmerick University College Dublin, Ireland November 10, 2004

4 ,

Andreas Heß ASSAM

slide-2
SLIDE 2

Outline Introduction ASSAM Annotator OATS Conclusion

Outline

1

Introduction Motivation Progress made

2

ASSAM Annotator The Annotator Application Relational Learning Evaluation

3

OATS

4

Conclusion

Andreas Heß ASSAM

slide-3
SLIDE 3

Outline Introduction ASSAM Annotator OATS Conclusion Motivation Progress made

Outline

1

Introduction Motivation Progress made

2

ASSAM Annotator The Annotator Application Relational Learning Evaluation

3

OATS

4

Conclusion

Andreas Heß ASSAM

slide-4
SLIDE 4

Outline Introduction ASSAM Annotator OATS Conclusion Motivation Progress made

Outline

1

Introduction Motivation Progress made

2

ASSAM Annotator The Annotator Application Relational Learning Evaluation

3

OATS

4

Conclusion

Andreas Heß ASSAM

slide-5
SLIDE 5

Outline Introduction ASSAM Annotator OATS Conclusion Motivation Progress made

Semantic Web Services

Desired Features Automatic discovery Automatic composition Automatic invocation

Andreas Heß ASSAM

slide-6
SLIDE 6

Outline Introduction ASSAM Annotator OATS Conclusion Motivation Progress made

Semantic Metadata

Assumptions Semantic annotation, e.g. OWL-S Shared ontology Problems Hand-crafting annotations can be tedious Integrate legacy web services Our Solution Learn mappings to ontology!

Andreas Heß ASSAM

slide-7
SLIDE 7

Outline Introduction ASSAM Annotator OATS Conclusion Motivation Progress made

Semantic Metadata

Assumptions Semantic annotation, e.g. OWL-S Shared ontology Problems Hand-crafting annotations can be tedious Integrate legacy web services Our Solution Learn mappings to ontology!

Andreas Heß ASSAM

slide-8
SLIDE 8

Outline Introduction ASSAM Annotator OATS Conclusion Motivation Progress made

Semantic Metadata

Assumptions Semantic annotation, e.g. OWL-S Shared ontology Problems Hand-crafting annotations can be tedious Integrate legacy web services Our Solution Learn mappings to ontology!

Andreas Heß ASSAM

slide-9
SLIDE 9

Outline Introduction ASSAM Annotator OATS Conclusion Motivation Progress made

What is learned?

Learn from: Web Service as a whole Operations Parameters Features used: Names Comments (if available)

Operations Parameters Web Service

Andreas Heß ASSAM

slide-10
SLIDE 10

Outline Introduction ASSAM Annotator OATS Conclusion Motivation Progress made

Outline

1

Introduction Motivation Progress made

2

ASSAM Annotator The Annotator Application Relational Learning Evaluation

3

OATS

4

Conclusion

Andreas Heß ASSAM

slide-11
SLIDE 11

Outline Introduction ASSAM Annotator OATS Conclusion Motivation Progress made

Progress

Earlier Classifying Services Ensemble Learning Now Classifying Services, Operations, Datatypes Relational Learning Application that can export OWL-S New OATS algorithm for

  • utput aggregation

Andreas Heß ASSAM

slide-12
SLIDE 12

Outline Introduction ASSAM Annotator OATS Conclusion Motivation Progress made

Progress

Earlier Classifying Services Ensemble Learning Now Classifying Services, Operations, Datatypes Relational Learning Application that can export OWL-S New OATS algorithm for

  • utput aggregation

Andreas Heß ASSAM

slide-13
SLIDE 13

Outline Introduction ASSAM Annotator OATS Conclusion Motivation Progress made

Progress

Earlier Classifying Services Ensemble Learning Now Classifying Services, Operations, Datatypes Relational Learning Application that can export OWL-S New OATS algorithm for

  • utput aggregation

Andreas Heß ASSAM

slide-14
SLIDE 14

Outline Introduction ASSAM Annotator OATS Conclusion Motivation Progress made

Progress

Earlier Classifying Services Ensemble Learning Now Classifying Services, Operations, Datatypes Relational Learning Application that can export OWL-S New OATS algorithm for

  • utput aggregation

Andreas Heß ASSAM

slide-15
SLIDE 15

Outline Introduction ASSAM Annotator OATS Conclusion Motivation Progress made

Progress

Earlier Classifying Services Ensemble Learning Now Classifying Services, Operations, Datatypes Relational Learning Application that can export OWL-S New OATS algorithm for

  • utput aggregation

Andreas Heß ASSAM

slide-16
SLIDE 16

Outline Introduction ASSAM Annotator OATS Conclusion Motivation Progress made

Progress

Earlier Classifying Services Ensemble Learning Now Classifying Services, Operations, Datatypes Relational Learning Application that can export OWL-S New OATS algorithm for

  • utput aggregation

Andreas Heß ASSAM

slide-17
SLIDE 17

Outline Introduction ASSAM Annotator OATS Conclusion The Annotator Application Relational Learning Evaluation

Outline

1

Introduction Motivation Progress made

2

ASSAM Annotator The Annotator Application Relational Learning Evaluation

3

OATS

4

Conclusion

Andreas Heß ASSAM

slide-18
SLIDE 18

Outline Introduction ASSAM Annotator OATS Conclusion The Annotator Application Relational Learning Evaluation

Outline

1

Introduction Motivation Progress made

2

ASSAM Annotator The Annotator Application Relational Learning Evaluation

3

OATS

4

Conclusion

Andreas Heß ASSAM

slide-19
SLIDE 19

Outline Introduction ASSAM Annotator OATS Conclusion The Annotator Application Relational Learning Evaluation

Our Application

Andreas Heß ASSAM

slide-20
SLIDE 20

Outline Introduction ASSAM Annotator OATS Conclusion The Annotator Application Relational Learning Evaluation

Our Application

(⌥⌧⌃⇢⌘⇧- ⌘⇤⌧⌘✓⌘⇢- ⇡/⌧ ⌘< =⌃@ ⌃⇧A/✏⌃ +⇧⌃⌃ A/⌃= ⌘⇤ ⌃⇧A/✏⌃ 2✓⌥/⇤ ?>*⇡ A/⌃= *⌘✏6⌃⇤⌧⌥⌧/⌘⇤ <⌘⇤⌅ /⇤ ?>*⇡ *⌘6⌥/⇤L*⌥⌧⌥⌧-%⌃ ⌘⇤⌧⌘✓⌘⇢/⌃

Andreas Heß ASSAM

slide-21
SLIDE 21

Outline Introduction ASSAM Annotator OATS Conclusion The Annotator Application Relational Learning Evaluation

ASSAM’s Key Feature

3⌃✏⌘66⌃⇤⌅⌃⌅ ⌥⇤⇤⌘⌧⌥⌧/⌘⇤

Andreas Heß ASSAM

slide-22
SLIDE 22

Outline Introduction ASSAM Annotator OATS Conclusion The Annotator Application Relational Learning Evaluation

OWL-S Export

!⇥⇤⌅⇧⌃⌅⌥ ⌦↵,⇧
.✏⇣⌘✓⌥◆✓⌦⇣5 !⌦↵,⇧6⌫⌃⌅⌥⇠, ⌦↵,⇧⌦✓⇥6⌦⇡✓✏⇣◆◆⇢⇧;;⌧⇥=6>◆?⌥@6⇡↵@?✓;⇥⇤⌅;.⌥◆⌥◆!⇢✓@⇥⇤⌅"⌘✓⌥◆✓⌦⇣;5 !;⇥⇤⌅⇧⌃⌅⌥5 !⇥⇤⌅⇧.⌥◆⌥◆!⇢✓#⌦⇥⇢✓⌦◆! ⌦↵,⇧
.✏⇣$?⌧✓⇣5 !⌦↵,⇧⌦⌥>=✓ ⌦↵,⇧⌦✓⇥6⌦⇡✓✏⇣◆◆⇢⇧;;⇤⇤⇤@⇤%@⇥⌦=;&''(;)*+,⇡✓⌧⌥"◆⌦?>=⇣;5 !⌦↵,⇧↵⇥⌧⌥?> ⌦↵,⇧⌦✓⇥6⌦⇡✓✏⇣"⌘✓⌥◆✓⌦⇣;5 !⌦↵,⇧6⌫#⌦⇥⇢✓⌦◆!⇠, ⌦↵,⇧⌦✓⇥6⌦⇡✓✏⇣◆◆⇢⇧;;⌧⇥=6>◆?⌥@6⇡↵@?✓;⇥⇤⌅;.⌥◆⌥◆!⇢✓@⇥⇤⌅"$?⌧✓⇣;5 !;⇥⇤⌅⇧.⌥◆⌥◆!⇢✓#⌦⇥⇢✓⌦◆!5 !⇥⇤⌅⇧.⌥◆⌥◆!⇢✓#⌦⇥⇢✓⌦◆! ⌦↵,⇧
.✏⇣$✓⌧⇢✓⌦⌥◆6⌦✓⇣5 !⌦↵,⇧⌦⌥>=✓ ⌦↵,⇧⌦✓⇥6⌦⇡✓✏⇣◆◆⇢⇧;;⇤⇤⇤@⇤%@⇥⌦=;&''(;)*+,⇡✓⌧⌥"◆⌦?>=⇣;5 !⌦↵,⇧↵⇥⌧⌥?> ⌦↵,⇧⌦✓⇥6⌦⇡✓✏⇣"⌘✓⌥◆✓⌦⇣;5 !⌦↵,⇧6⌫#⌦⇥⇢✓⌦◆!⇠, ⌦↵,⇧⌦✓⇥6⌦⇡✓✏⇣◆◆⇢⇧;;⌧⇥=6>◆?⌥@6⇡↵@?✓;⇥⇤⌅;.⌥◆⌥◆!⇢✓@⇥⇤⌅"$✓⌧⇢✓⌦⌥◆6⌦✓⇣;5 !;⇥⇤⌅⇧.⌥◆⌥◆!⇢✓#⌦⇥⇢✓⌦◆!5 !=⌦⇥6>↵?>=⇧⇤↵⌅⇠6◆⇢6◆*✓⌥=✓*⌥⇢5 !=⌦⇥6>↵?>=⇧⇤↵⌅*✓⌥=✓#⌥⌦◆5 !-↵⇧⌥>!./
 ⌦↵,⇧0⌥⌅6✓✏⇣1◆✓2⇤↵⌅3"4⇥↵!⇣;5 !;=⌦⇥6>↵?>=⇧⇤↵⌅*✓⌥=✓#⌥⌦◆5 !=⌦⇥6>↵?>=⇧-⌅◆$⌦⌥>,⇥⌦⌧⌥◆?⇥> ⌦↵,⇧⇢⌥⌦✓$!⇢✓✏⇣+?◆✓⌦⌥⌅⇣5 !-⌅⇧◆!⌅✓✓✓◆ 0✓⌦?⇥>✏⇣(@'⇣

  • ⌧⌅>⇧-⌅✏⇣◆◆⇢⇧;;⇤⇤⇤@⇤%@⇥⌦=;(555;),+;$⌦⌥>,⇥⌦⌧⇣5

!-⌅⇧⇥6◆⇢6◆ ⌧✓◆⇥↵✏⇣-⌧⌅⇣ ?>↵✓>◆✏⇣!✓⇣;5 !-⌅⇧◆✓⌧⇢⌅⌥◆✓ ⌧⌥◆⇡✏⇣;⇣5 !⌦↵,⇧/.6 -⌧⌅>⇧⌦↵,✏⇣◆◆⇢⇧;;⇤⇤⇤@⇤%@⇥⌦=;(555;'&;&&7⌦↵,7!>◆⌥-7>⇣

  • ⌧⌅>⇧◆✓2⇡⇥>⇡✓⇢◆✏⇣1◆✓2⇡⇥>⇡✓⇢◆3⇣
  • ⌧⌅>⇧◆✓2⇢⌦⇥⇡✓✏⇣1◆✓2⇢⌦⇥⇡✓3⇣5

!◆✓2⇡⇥>⇡✓⇢◆⇧⌘✓⌥◆✓⌦5 !◆✓2⇡⇥>⇡✓⇢◆⇧$?⌧✓5 !-⌅⇧0⌥⌅6✓7⇥, ✓⌅✓⇡◆✏⇣4⇥↵!;$?⌧✓⇣;5 !;◆✓2⇡⇥>⇡✓⇢◆⇧$?⌧✓5

Andreas Heß ASSAM

slide-23
SLIDE 23

Outline Introduction ASSAM Annotator OATS Conclusion The Annotator Application Relational Learning Evaluation

OWL-S Export

ASSAM generates... Grounding Profile Process Model Concepts (like WSDL2DAML-S by Paolucci, Sycara et al.)

Andreas Heß ASSAM

slide-24
SLIDE 24

Outline Introduction ASSAM Annotator OATS Conclusion The Annotator Application Relational Learning Evaluation

OWL-S Export – Additional Features

But ASSAM also... generates XSLT transformation for Grounding uses shared ontologies for Concepts and Profile

Andreas Heß ASSAM

slide-25
SLIDE 25

Outline Introduction ASSAM Annotator OATS Conclusion The Annotator Application Relational Learning Evaluation

Outline

1

Introduction Motivation Progress made

2

ASSAM Annotator The Annotator Application Relational Learning Evaluation

3

OATS

4

Conclusion

Andreas Heß ASSAM

slide-26
SLIDE 26

Outline Introduction ASSAM Annotator OATS Conclusion The Annotator Application Relational Learning Evaluation

Definitions

Category Description of service as a whole e.g. Weather, Finance, Books Profile Domain Purpose of single operation e.g. Query Price, Purchase Book PEs Datatype Meaning of single parameter e.g. Title, Author’s Name IOs

5

Andreas Heß ASSAM

slide-27
SLIDE 27

Outline Introduction ASSAM Annotator OATS Conclusion The Annotator Application Relational Learning Evaluation

Definitions: What? Why?

No new standard! We do not propose a new standard or ontology language Concepts independent of language

Andreas Heß ASSAM

slide-28
SLIDE 28

Outline Introduction ASSAM Annotator OATS Conclusion The Annotator Application Relational Learning Evaluation

Relations: Example

/

.⌘⌘7 (⌥⌧⌃⇢⌘⇧- 1⌃⇧- @⌘⌘7 %⇧/✏⌃ *⌘6⌥/⇤ *⌥⌧⌥⌧-%⌃ .⌘⌘7 ⌧/⌧✓⌃

Andreas Heß ASSAM

slide-29
SLIDE 29

Outline Introduction ASSAM Annotator OATS Conclusion The Annotator Application Relational Learning Evaluation

Relations: Example

+⌃⌥ (⌥⌧⌃⇢⌘⇧- *⌘6⌥/⇤ *⌥⌧⌥⌧-%⌃ $⇧⌅⌃⇧ ⌧⌃⌥

⇥

.⌘⌘7 ⌧/⌧✓⌃//

Andreas Heß ASSAM

slide-30
SLIDE 30

Outline Introduction ASSAM Annotator OATS Conclusion The Annotator Application Relational Learning Evaluation

Relations: Example

+⌃⌥ (⌥⌧⌃⇢⌘⇧- *⌘6⌥/⇤ *⌥⌧⌥⌧-%⌃ $⇧⌅⌃⇧ ⌧⌃⌥

⇥

(⇧⌃⌅/⌧ ✏⌥⇧⌅ ⇤6@⌃⇧

Andreas Heß ASSAM

slide-31
SLIDE 31

Outline Introduction ASSAM Annotator OATS Conclusion The Annotator Application Relational Learning Evaluation

Relations: Example

(⌥⌧⌃⇢⌘⇧- *⌘6⌥/⇤ *⌥⌧⌥⌧-%⌃ .⌘⌘7 ⌧/⌧✓⌃

/

Andreas Heß ASSAM

slide-32
SLIDE 32

Outline Introduction ASSAM Annotator OATS Conclusion The Annotator Application Relational Learning Evaluation

Relations: Example

/

(⌥⌧⌃⇢⌘⇧- 1⌃⇧- @⌘⌘7 %⇧/✏⌃ *⌘6⌥/⇤ *⌥⌧⌥⌧-%⌃ .⌘⌘7 ⌧/⌧✓⌃

/

Andreas Heß ASSAM

slide-33
SLIDE 33

Outline Introduction ASSAM Annotator OATS Conclusion The Annotator Application Relational Learning Evaluation

Relations: Example

/

.⌘⌘7 (⌥⌧⌃⇢⌘⇧- 1⌃⇧- @⌘⌘7 %⇧/✏⌃ *⌘6⌥/⇤ *⌥⌧⌥⌧-%⌃ .⌘⌘7 ⌧/⌧✓⌃

Andreas Heß ASSAM

slide-34
SLIDE 34

Outline Introduction ASSAM Annotator OATS Conclusion The Annotator Application Relational Learning Evaluation

Relational Learning

Ways to exploit relations Bayesian Networks (see our earlier paper) Iterative Classification

Andreas Heß ASSAM

slide-35
SLIDE 35

Outline Introduction ASSAM Annotator OATS Conclusion The Annotator Application Relational Learning Evaluation

Iterative Classification – Example

Classification in round n based on classification from round n − 1 Category initially unknown Domain initially unknown Datatype initially unknown

Andreas Heß ASSAM

slide-36
SLIDE 36

Outline Introduction ASSAM Annotator OATS Conclusion The Annotator Application Relational Learning Evaluation

Iterative Classification – Example

Classification in round 1 is made in standard way Category Communication Domain Query tea price Datatype Person’s name

Andreas Heß ASSAM

slide-37
SLIDE 37

Outline Introduction ASSAM Annotator OATS Conclusion The Annotator Application Relational Learning Evaluation

Iterative Classification – Example

Classification in round 2 is based on classifcation from round 1 Category Communication Tea Domain Query tea price Query book price Datatype Person’s name Sender’s name

Andreas Heß ASSAM

slide-38
SLIDE 38

Outline Introduction ASSAM Annotator OATS Conclusion The Annotator Application Relational Learning Evaluation

Iterative Classification – Example

Classification in round 3 is based on classifcation from round 2 Category Communication Tea Books Domain Query tea price Query book price Query book price Datatype Person’s name Sender’s name Author’s name

Andreas Heß ASSAM

slide-39
SLIDE 39

Outline Introduction ASSAM Annotator OATS Conclusion The Annotator Application Relational Learning Evaluation

Iterative Classification – Example

Observation Iterative classification can correct initial errors! Category Communication Tea Books Domain Query tea price Query book price Query book price Datatype Person’s name Sender’s name Author’s name

Andreas Heß ASSAM

slide-40
SLIDE 40

Outline Introduction ASSAM Annotator OATS Conclusion The Annotator Application Relational Learning Evaluation

Two Views

Intrinsic View Features inherent to instance e.g. name of operation Extrinsic View Relations between instances e.g. annotation of parameters (Following Neville and Jensen)

Andreas Heß ASSAM

slide-41
SLIDE 41

Outline Introduction ASSAM Annotator OATS Conclusion The Annotator Application Relational Learning Evaluation

Two Views

Intrinsic View Features inherent to instance e.g. name of operation Extrinsic View Relations between instances e.g. annotation of parameters (Following Neville and Jensen)

Andreas Heß ASSAM

slide-42
SLIDE 42

Outline Introduction ASSAM Annotator OATS Conclusion The Annotator Application Relational Learning Evaluation

Iterative Algorithm for Relational Classification

Iterative Algorithm Features are combination

  • f intrinsic and extrinsic

Extrinsic view changes with new results Predictions are fed back Extrinsic view is dynamic

intrinsic predictions extrinsic/dynamic intrinsic predictions

e.g. Neville/Jensen, Chakrabarti, Lu/Getoor

Andreas Heß ASSAM

slide-43
SLIDE 43

Outline Introduction ASSAM Annotator OATS Conclusion The Annotator Application Relational Learning Evaluation

Iterative Ensemble Algorithm for Relational Classification

Iterative Ensemble Algorithm Separate intrinsic/extrinsic classifiers Better performance with high-dimensional features

intrinsic extrinsic/dynamic intrinsic predictions predictions

Andreas Heß ASSAM

slide-44
SLIDE 44

Outline Introduction ASSAM Annotator OATS Conclusion The Annotator Application Relational Learning Evaluation

Iterative Ensemble Algorithm with Specialised Classifiers

Specialised Classifiers Train several classifiers on intrinsic features Each one trained on subset of instances Classifier selected based

  • n extrinsic features

intrinsic intrinsic predictions predictions 2 2 select

Andreas Heß ASSAM

slide-45
SLIDE 45

Outline Introduction ASSAM Annotator OATS Conclusion The Annotator Application Relational Learning Evaluation

Specialised Classifiers: Example

intrinsic intrinsic predictions 2 select predictions

Andreas Heß ASSAM

slide-46
SLIDE 46

Outline Introduction ASSAM Annotator OATS Conclusion The Annotator Application Relational Learning Evaluation

Specialised Classifiers: Example

intrinsic intrinsic predictions 2 select

Andreas Heß ASSAM

slide-47
SLIDE 47

Outline Introduction ASSAM Annotator OATS Conclusion The Annotator Application Relational Learning Evaluation

Specialised Classifiers

Specialised Classifiers Useful if extrinsic view not sufficient on its own Extrinsic features serve as selector Idea can be applied in various ways

Andreas Heß ASSAM

slide-48
SLIDE 48

Outline Introduction ASSAM Annotator OATS Conclusion The Annotator Application Relational Learning Evaluation

Specialised Classifiers

Specialised Classifiers Useful if extrinsic view not sufficient on its own Extrinsic features serve as selector Idea can be applied in various ways

Andreas Heß ASSAM

slide-49
SLIDE 49

Outline Introduction ASSAM Annotator OATS Conclusion The Annotator Application Relational Learning Evaluation

Specialised Classifiers

Specialised Classifiers Useful if extrinsic view not sufficient on its own Extrinsic features serve as selector Idea can be applied in various ways

Andreas Heß ASSAM

slide-50
SLIDE 50

Outline Introduction ASSAM Annotator OATS Conclusion The Annotator Application Relational Learning Evaluation

Outline

1

Introduction Motivation Progress made

2

ASSAM Annotator The Annotator Application Relational Learning Evaluation

3

OATS

4

Conclusion

Andreas Heß ASSAM

slide-51
SLIDE 51

Outline Introduction ASSAM Annotator OATS Conclusion The Annotator Application Relational Learning Evaluation

Dataset

Our Web Services Dataset 164 Web Services in 22 Categories 1138 Operations in 136 Domains 5452 Parameters with 312 Datatypes Download This collection is available for download from the Repository of Semantic Web Services: http://smi.ucd.ie/RSWS

Andreas Heß ASSAM

slide-52
SLIDE 52

Outline Introduction ASSAM Annotator OATS Conclusion The Annotator Application Relational Learning Evaluation

Setup

Experimental Setup Using iterative ensemble for service, operations Using specialised classifiers for datatypes Leave-one-service-out Compared to omniscient setup with extrinsic view always correct Compared to non-relational baseline Non-ensemble setting ommitted: always worse than baseline

Andreas Heß ASSAM

slide-53
SLIDE 53

Outline Introduction ASSAM Annotator OATS Conclusion The Annotator Application Relational Learning Evaluation

Evaluation: Datatype of Parameters

20 30 40 50 60 70 1 2 3 Accuracy % Tolerance Omniscient Specialised Baseline

Andreas Heß ASSAM

slide-54
SLIDE 54

Outline Introduction ASSAM Annotator OATS Conclusion The Annotator Application Relational Learning Evaluation

Evaluation: Domain of Operations

20 30 40 50 60 1 2 3 Accuracy % Tolerance Omniscient Iterative Ensemble Baseline

Andreas Heß ASSAM

slide-55
SLIDE 55

Outline Introduction ASSAM Annotator OATS Conclusion The Annotator Application Relational Learning Evaluation

Evaluation: Category of Service

60 70 80 90 1 2 3 Accuracy % Tolerance Omniscient Iterative Ensemble Baseline

Andreas Heß ASSAM

slide-56
SLIDE 56

Outline Introduction ASSAM Annotator OATS Conclusion

Outline

1

Introduction Motivation Progress made

2

ASSAM Annotator The Annotator Application Relational Learning Evaluation

3

OATS

4

Conclusion

Andreas Heß ASSAM

slide-57
SLIDE 57

Outline Introduction ASSAM Annotator OATS Conclusion

Operation Aggregation Tool for Web Services

Key Features Aggregates output from (legacy) Web Services Considers instances data Uses probes to influence output Use string distance metrics to match outputs

Andreas Heß ASSAM

slide-58
SLIDE 58

Outline Introduction ASSAM Annotator OATS Conclusion

ASSAM and OATS

ASSAM Complete Service Ignores instance data Annotation OATS Output only Uses instance data Aggregation

Andreas Heß ASSAM

slide-59
SLIDE 59

Outline Introduction ASSAM Annotator OATS Conclusion

Active Probing & Matching Outputs

6&()7&' )&8% /() .95)5 7*.7 /+6 :;<:=> ?@-8%7-A>B CD =C 04(5) )8(E 6,F5%F > )8*, CD >GH-??-8%7 :;<:=I=: =;

O2 O1

JI?D= :;<:=K?DD<L $&())/&

Andreas Heß ASSAM

slide-60
SLIDE 60

Outline Introduction ASSAM Annotator OATS Conclusion

Ensemble of String Distance Metrics

04(5) )6(7 8,95%9 : )6*, ;; :<=->>-6%? @AB@C;C@ CA <9*)D9*5)(,4&-2(5&9-EF&3&,5)?&*,G #+H&,-2(5&9-I-JK2'*9-E#LMNLI-$+0)#LMNLG <9*)D9*5)(,4&-2(5&9-EO('+D1*,H/&'G

Andreas Heß ASSAM

slide-61
SLIDE 61

Outline Introduction ASSAM Annotator OATS Conclusion

Multiple Probes

Multiple probes minimize the effect of spurious data

*+,'(+% '+)- .','+ (&#( "$* *+,'(+% ./0 *&12 '+)- '+)- 3&'0 "$*

4""&1$&. 5+*67$%/ 8,"&9$%1&, 5+*6:+%.+0 8(&3,#$ 5+*7$%/ ;,16:$.+ :+%.+0 <= 1$2,', 1$2,', 1$2,', <= >< ?> >? <@ >A ?B >? <= >< ?> >? 9,&% 1C, .D110 9,&% EF6G >F65 =F6H <F65I

GM G< G>

Andreas Heß ASSAM

slide-62
SLIDE 62

Outline Introduction ASSAM Annotator OATS Conclusion

OATS – Algorithm

The OATS Algorithm Compute pairwise distance between outputs Perform HAC clustering Match closest pairs

Andreas Heß ASSAM

slide-63
SLIDE 63

Outline Introduction ASSAM Annotator OATS Conclusion

OATS – Advanced

Advanded Features Learning weights for ensemble For more details... ... see Eddie Johnston, EKAW 2004

Andreas Heß ASSAM

slide-64
SLIDE 64

Outline Introduction ASSAM Annotator OATS Conclusion

OATS – Advanced

Advanded Features Learning weights for ensemble For more details... ... see Eddie Johnston, EKAW 2004

Andreas Heß ASSAM

slide-65
SLIDE 65

Outline Introduction ASSAM Annotator OATS Conclusion

Outline

1

Introduction Motivation Progress made

2

ASSAM Annotator The Annotator Application Relational Learning Evaluation

3

OATS

4

Conclusion

Andreas Heß ASSAM

slide-66
SLIDE 66

Outline Introduction ASSAM Annotator OATS Conclusion

Summary

We have presented: Relational Ensemble Learning for Web Service Annotation ASSAM Annotator Application OATS Algorithm for output aggregation

Andreas Heß ASSAM

slide-67
SLIDE 67

Outline Introduction ASSAM Annotator OATS Conclusion

The Bigger Picture

ASSAM’s Limitations Supervised Learning: Need training data Pieces in the Puzzle Create and/or select domain ontologies Process composition ...

Andreas Heß ASSAM

slide-68
SLIDE 68

Outline Introduction ASSAM Annotator OATS Conclusion

Discussion

Thank You for Your Attention Questions? URLs http://moguntia.ucd.ie/projects/annotator/ http://smi.ucd.ie/RSWS Demo Taste ASSAM in the Demo session: 17.00-18.30, Rooms 5 & 6

Andreas Heß ASSAM