SlideShare a Scribd company logo
1 of 33
Download to read offline
Introduction Three classiļ¬ers Experiments and results Summary
Similarity Features, and their Role in Concept
Alignment Learning
Shenghui Wang1 Gwenn Englebienne2 Christophe GuĀ“eret 1
Stefan Schlobach1 Antoine Isaac1 Martijn Schut1
1 Vrije Universiteit Amsterdam
2 Universiteit van Amsterdam
SEMAPRO 2010
Florence
Introduction Three classiļ¬ers Experiments and results Summary
Outline
1 Introduction
Classiļ¬cation of concept mappings based on instance
similarity
2 Three classiļ¬ers
Markov Random Field
Multi-objective Evolution Strategy
Support Vector Machine
3 Experiments and results
4 Summary
Introduction Three classiļ¬ers Experiments and results Summary
Thesaurus mapping
SemanTic Interoperability To access Cultural Heritage
(STITCH) through mappings between thesauri
Scope of the problem:
Big thesauri with tens of thousands of concepts
Huge collections (e.g., National Library of the Neterlands:
80km of books in one collection)
Heterogeneous (e.g., books, manuscripts, illustrations, etc.)
Multi-lingual problem
Solving matching problems is one step to the solution of the
interoperability problem.
e.g., ā€œplankzeilenā€ vs. ā€œsurfsportā€
e.g., ā€œarcheologyā€ vs. ā€œexcavationā€
Introduction Three classiļ¬ers Experiments and results Summary
Thesaurus mapping
SemanTic Interoperability To access Cultural Heritage
(STITCH) through mappings between thesauri
Scope of the problem:
Big thesauri with tens of thousands of concepts
Huge collections (e.g., National Library of the Neterlands:
80km of books in one collection)
Heterogeneous (e.g., books, manuscripts, illustrations, etc.)
Multi-lingual problem
Solving matching problems is one step to the solution of the
interoperability problem.
e.g., ā€œplankzeilenā€ vs. ā€œsurfsportā€
e.g., ā€œarcheologyā€ vs. ā€œexcavationā€
Introduction Three classiļ¬ers Experiments and results Summary
Thesaurus mapping
SemanTic Interoperability To access Cultural Heritage
(STITCH) through mappings between thesauri
Scope of the problem:
Big thesauri with tens of thousands of concepts
Huge collections (e.g., National Library of the Neterlands:
80km of books in one collection)
Heterogeneous (e.g., books, manuscripts, illustrations, etc.)
Multi-lingual problem
Solving matching problems is one step to the solution of the
interoperability problem.
e.g., ā€œplankzeilenā€ vs. ā€œsurfsportā€
e.g., ā€œarcheologyā€ vs. ā€œexcavationā€
Introduction Three classiļ¬ers Experiments and results Summary
Automatic alignment techniques
Lexical
labels and textual information of entities
Structural
structure of the formal deļ¬nitions of entities, position in the
hierarchy
Extensional
statistical information of instances, i.e., objects indexed with
entities
Background knowledge
using a shared conceptual reference to ļ¬nd links indirectly
Introduction Three classiļ¬ers Experiments and results Summary
Instance-based techniques: common instance based
Introduction Three classiļ¬ers Experiments and results Summary
Instance-based techniques: common instance based
Introduction Three classiļ¬ers Experiments and results Summary
Instance-based techniques: common instance based
Introduction Three classiļ¬ers Experiments and results Summary
Pros and cons
Advantages
Simple to implement
Interesting results
Disadvantages
Requires suļ¬ƒcient amounts of common instances
Only uses part of the available information
Introduction Three classiļ¬ers Experiments and results Summary
Instance-based techniques: Instance similarity based
Introduction Three classiļ¬ers Experiments and results Summary
Instance-based techniques: Instance similarity based
Introduction Three classiļ¬ers Experiments and results Summary
Instance-based techniques: Instance similarity based
Introduction Three classiļ¬ers Experiments and results Summary
Representing concepts and the similarity between them
Instance features Concept features Pair features
Cos. dist.
Bag of words
Bag of words
Bag of words
Bag of words
Bag of words
Bag of words
Creator
Title
Publisher
...
Creator
Title
Publisher
...
Creator
Title
Publisher
...
...
f1
f2
f3
Concept1Concept2
{{
{
{
{
Creator
Title
Publisher
...
Creator
Title
Publisher
...
Creator
Term 1: 4
Term 2: 1
Term 3: 0
...
Title
Term 1: 0
Term 2: 3
Term 3: 0
...
Publisher
Term 1: 2
Term 2: 1
Term 3: 3
...
Creator
Term 1: 2
Term 2: 0
Term 3: 0
...
Title
Term 1: 0
Term 2: 4
Term 3: 1
...
Publisher
Term 1: 4
Term 2: 1
Term 3: 1
...
Cos. dist.
Cos. dist.
Introduction Three classiļ¬ers Experiments and results Summary
Classiļ¬cation of concept mappings based on instance similarity
Classiļ¬cation based on instance similarity
Each pair of concepts is treated as a point in a ā€œsimilarity
spaceā€
Its position is deļ¬ned by the features of the pair.
The features of the pair are the diļ¬€erent measures of similarity
between the conceptsā€™ instances.
Hypothesis: the label of a point ā€” which represents whether
the pair is a positive mapping or negative one ā€” is correlated
with the position of this point in this space.
With already labelled points and the actual similarity values of
concepts involved, it is possible to classify a point, i.e., to give
it a right label, based on its location given by the actual
similarity values.
Introduction Three classiļ¬ers Experiments and results Summary
Classiļ¬cation of concept mappings based on instance similarity
Research questions
How do diļ¬€erent classiļ¬ers perform on this instance-based
mapping task?
What are the beneļ¬ts of using a machine learning algorithm
to determine the importance of features?
Are there regularities wrt. the relative importance given to
speciļ¬c features for similarity computation? Are these weights
related to application data characteristics?
Introduction Three classiļ¬ers Experiments and results Summary
Three classiļ¬ers used
Markov Random Field (MRF)
Evolutionary Strategy (ES)
Support Vector Machine (SVM)
Introduction Three classiļ¬ers Experiments and results Summary
Markov Random Field
Markov Random Field
Let T = { (x(i), y(i)) }N
i=1 be the training set
x(i)
āˆˆ RK
, the features
y(i)
āˆˆ Y = {positive, negative}, the label
The conditional probability of a label given the input is
modelled as
p(y(i)
|xi , Īø) =
1
Z(xi , Īø)
exp
K
j=1
Ī»j Ļ†j (y(i)
, x(i)
) , (1)
where Īø = { Ī»j }K
j=1 are the weights associated to the feature
functions Ļ† and Z(xi , Īø) is a normalisation constant
Introduction Three classiļ¬ers Experiments and results Summary
Markov Random Field
The classiļ¬er used: Markov Random Field (contā€™)
The likelihood of the data set for given model parameters
p(T|Īø) is given by:
p(T|Īø) =
N
i=1
p(y(i)
|x(i)
) (2)
During learning, our objective is to ļ¬nd the most likely values
for Īø for the given training data.
The decision criterion for assigning a label y(i) to a new pair
of concepts i is then simply given by:
y(i)
= argmax
y
p(y|x(i)
) (3)
Introduction Three classiļ¬ers Experiments and results Summary
Multi-objective Evolution Strategy
Multi-objective Evolution Strategy
Evolutionary strategies (ES) have two characteristic properties:
ļ¬rstly, they are used for continuous value optimisation, and,
secondly, they are self-adaptive.
An ES individual is a direct model of the searched solution,
deļ¬ned by Ī› and some evolution strategy parameters:
Ī›, Ī£ ā†” Ī»1, . . . , Ī»K , Ļƒ1, . . . , ĻƒK (4)
The ļ¬tness function is related to the decision criterion for the
ES, which is sign-based:
LES
i =
1 if K
j=1 Ī»i Fij > 0
0 otherwise
(5)
Introduction Three classiļ¬ers Experiments and results Summary
Multi-objective Evolution Strategy
Multi-objective Evolution Strategy (contā€™)
Maximising the number of positive results and negative results
are two opposite goals.
f1(Ī› | F, L) = #{Fi |
K
j=1
Ī»i Fij > 0 āˆ§ Li = 1} (6)
f2(Ī› | F, L) = #{Fi |
K
j=1
Ī»i Fij ā‰¤ 0 āˆ§ Li = 0} (7)
Introduction Three classiļ¬ers Experiments and results Summary
Multi-objective Evolution Strategy
Multi-objective Evolution Strategy (contā€™)
Evolution process
Recombination: Two parent individuals are combined using
diļ¬€erent weighting, producing two new individuals
Mutation: One parent individual changes itself into a new child
individual
Survivor selection: NSGA-II
Introduction Three classiļ¬ers Experiments and results Summary
Support Vector Machine
Support Vector Machine
Support Vector Machine (SVM) is used as a maximum margin
classiļ¬er whose task consists in ļ¬nding an hyperplane separating
the two classes.
The objective is to maximise the margin separating the two
classes whilst minimizing classiļ¬cation error risk.
Introduction Three classiļ¬ers Experiments and results Summary
Experiments and results
Thesauri to match: GTT (35K) and Brinkman (5K)
Instances: 1 million books
GTT annotated books: 307K
Brinkman annotated books: 490K
Dually annotated books: 222K
Introduction Three classiļ¬ers Experiments and results Summary
Feature slection for similarity calculation
Ī»j Feature
1 Lexical
2 Jaccard
3 Date
4 ISBN
5 NBN
6 PPN
7 SelSleutel
8 abstract
9 alternative
10 annotation
Ī»j Feature
11 author
12 contributor
13 creator
14 dateCopyrighted
15 description
16 extent
17 hasFormat
18 hasPart
19 identiļ¬er
20 isVersionOf
Ī»j Feature
21 issued
22 language
23 mods:edition
24 publisher
25 refNBN
26 relation
27 spatial
28 subject
29 temporal
30 title
Table: List of the features
Introduction Three classiļ¬ers Experiments and results Summary
Quality of learning
0
0.2
0.4
0.6
0.8
1
Precision Recall F-measure
MRF 1-30
MRF 3-30
ES
SVM
0
0.2
0.4
0.6
0.8
1
Precision Recall F-measure
0
0.2
0.4
0.6
0.8
1
Precision Recall F-measure
0
0.2
0.4
0.6
0.8
1
Precision Recall F-measure
Figure: Precision, recall and F-Measure for mappings with a positive
label (top) and a negative label (bottom). Error bars indicate one
standard deviation over the 10 folds of cross-validation.
Introduction Three classiļ¬ers Experiments and results Summary
Relative importance of features
Which features of our instances are important for mapping?
Figure: Mutual information between features and labels
Introduction Three classiļ¬ers Experiments and results Summary
Relative importance of features
ES lambdas are not really conclusive
ES lambdas that are most inconclusive correspond to the least
informative features
Introduction Three classiļ¬ers Experiments and results Summary
Relative importance of features
Important features in terms of mutual information are
associated to large MRF weights
Introduction Three classiļ¬ers Experiments and results Summary
A more detailed analysis
Expected important features:
Label similarity (1), instance overlap (2), subject (28), etc.
Expected unimportant features:
Size of the book (16), format description (17) and language
(22), etc.
Surprisingly important features:
Date (14)
Surprisingly unimportant features:
Description (15) and abstract (8)
Introduction Three classiļ¬ers Experiments and results Summary
A more detailed analysis
Expected important features:
Label similarity (1), instance overlap (2), subject (28), etc.
Expected unimportant features:
Size of the book (16), format description (17) and language
(22), etc.
Surprisingly important features:
Date (14)
Surprisingly unimportant features:
Description (15) and abstract (8)
Introduction Three classiļ¬ers Experiments and results Summary
Summary
We tried three machine learning classiļ¬ers on the
instance-based mapping task, among which MRF and ES can
automatically identify meaningful features.
The MRF and the ES, result in a performance in the
neighbourhood of 90%, showing the validity of the approach.
Our analysis suggests that when many diļ¬€erent description
features interact, there is no systematic correlation between
what a learning method could ļ¬nd and what an application
expert may anticipate.
Introduction Three classiļ¬ers Experiments and results Summary
Thank you

More Related Content

What's hot

Machine learning and Neural Networks
Machine learning and Neural NetworksMachine learning and Neural Networks
Machine learning and Neural Networks
butest
Ā 
Lecture 2
Lecture 2Lecture 2
Lecture 2
butest
Ā 
The science behind predictive analytics a text mining perspective
The science behind predictive analytics  a text mining perspectiveThe science behind predictive analytics  a text mining perspective
The science behind predictive analytics a text mining perspective
ankurpandeyinfo
Ā 
Machine Learning: An Introduction Fu Chang
Machine Learning: An Introduction Fu ChangMachine Learning: An Introduction Fu Chang
Machine Learning: An Introduction Fu Chang
butest
Ā 

What's hot (18)

Discovering Novel Information with sentence Level clustering From Multi-docu...
Discovering Novel Information with sentence Level clustering  From Multi-docu...Discovering Novel Information with sentence Level clustering  From Multi-docu...
Discovering Novel Information with sentence Level clustering From Multi-docu...
Ā 
Biehl hanze-2021
Biehl hanze-2021Biehl hanze-2021
Biehl hanze-2021
Ā 
Towards Transfer Learning of Link Specifications
Towards Transfer Learning of Link SpecificationsTowards Transfer Learning of Link Specifications
Towards Transfer Learning of Link Specifications
Ā 
Machine learning and Neural Networks
Machine learning and Neural NetworksMachine learning and Neural Networks
Machine learning and Neural Networks
Ā 
Textmining Predictive Models
Textmining Predictive ModelsTextmining Predictive Models
Textmining Predictive Models
Ā 
Lecture 2
Lecture 2Lecture 2
Lecture 2
Ā 
Kernel methods for data integration in systems biology
Kernel methods for data integration in systems biologyKernel methods for data integration in systems biology
Kernel methods for data integration in systems biology
Ā 
Capter10 cluster basic
Capter10 cluster basicCapter10 cluster basic
Capter10 cluster basic
Ā 
The science behind predictive analytics a text mining perspective
The science behind predictive analytics  a text mining perspectiveThe science behind predictive analytics  a text mining perspective
The science behind predictive analytics a text mining perspective
Ā 
2.7 other classifiers
2.7 other classifiers2.7 other classifiers
2.7 other classifiers
Ā 
10 clusbasic
10 clusbasic10 clusbasic
10 clusbasic
Ā 
Dbm630 lecture09
Dbm630 lecture09Dbm630 lecture09
Dbm630 lecture09
Ā 
Machine Learning: An Introduction Fu Chang
Machine Learning: An Introduction Fu ChangMachine Learning: An Introduction Fu Chang
Machine Learning: An Introduction Fu Chang
Ā 
3.1 clustering
3.1 clustering3.1 clustering
3.1 clustering
Ā 
Investigating the 3D structure of the genome with Hi-C data analysis
Investigating the 3D structure of the genome with Hi-C data analysisInvestigating the 3D structure of the genome with Hi-C data analysis
Investigating the 3D structure of the genome with Hi-C data analysis
Ā 
10 clusbasic
10 clusbasic10 clusbasic
10 clusbasic
Ā 
Chapter 11 cluster advanced : web and text mining
Chapter 11 cluster advanced : web and text miningChapter 11 cluster advanced : web and text mining
Chapter 11 cluster advanced : web and text mining
Ā 
Kernel methods for data integration in systems biology
Kernel methods for data integration in systems biology Kernel methods for data integration in systems biology
Kernel methods for data integration in systems biology
Ā 

Viewers also liked

ss10_unit3
ss10_unit3ss10_unit3
ss10_unit3
Mr. Park
Ā 
Week 2
Week 2Week 2
Week 2
emankl
Ā 
7 foreign policy process (1)
7 foreign policy process (1)7 foreign policy process (1)
7 foreign policy process (1)
Ayesha Bhatti
Ā 
Multinational corporations
Multinational corporationsMultinational corporations
Multinational corporations
SHANTANU TYAGI
Ā 

Viewers also liked (10)

Alexander Bogomolov. RUSSIAN SOFT POWER IN UKRAINE: sources and effects
Alexander Bogomolov. RUSSIAN SOFT POWER IN UKRAINE: sources and effects Alexander Bogomolov. RUSSIAN SOFT POWER IN UKRAINE: sources and effects
Alexander Bogomolov. RUSSIAN SOFT POWER IN UKRAINE: sources and effects
Ā 
ss10_unit3
ss10_unit3ss10_unit3
ss10_unit3
Ā 
Week 2
Week 2Week 2
Week 2
Ā 
A study of the contemporary world:Pakistan
A study of the contemporary world:PakistanA study of the contemporary world:Pakistan
A study of the contemporary world:Pakistan
Ā 
The history and evolution of foreign policy analysis
The history and evolution of foreign policy analysisThe history and evolution of foreign policy analysis
The history and evolution of foreign policy analysis
Ā 
7 foreign policy process (1)
7 foreign policy process (1)7 foreign policy process (1)
7 foreign policy process (1)
Ā 
Foreign Policy and Level of Analysis Problem
Foreign Policy and Level of Analysis ProblemForeign Policy and Level of Analysis Problem
Foreign Policy and Level of Analysis Problem
Ā 
FOREIGN POLICY ANALYSIS
FOREIGN POLICY ANALYSISFOREIGN POLICY ANALYSIS
FOREIGN POLICY ANALYSIS
Ā 
Multinational corporations
Multinational corporationsMultinational corporations
Multinational corporations
Ā 
Introduction to political science
Introduction to political scienceIntroduction to political science
Introduction to political science
Ā 

Similar to Similarity Features, and their Role in Concept Alignment Learning

Tdm probabilistic models (part 2)
Tdm probabilistic  models (part  2)Tdm probabilistic  models (part  2)
Tdm probabilistic models (part 2)
KU Leuven
Ā 
Text categorization as a graph
Text categorization as a graphText categorization as a graph
Text categorization as a graph
James Wong
Ā 
Text categorization as a graph
Text categorization as a graphText categorization as a graph
Text categorization as a graph
Young Alista
Ā 
Text categorization as a graph
Text categorization as a graphText categorization as a graph
Text categorization as a graph
Fraboni Ec
Ā 
Text categorization
Text categorization Text categorization
Text categorization
Luis Goldster
Ā 
Text categorization as a graph
Text categorization as a graph Text categorization as a graph
Text categorization as a graph
David Hoen
Ā 
Online Multi-Person Tracking Using Variance Magnitude of Image colors and Sol...
Online Multi-Person Tracking Using Variance Magnitude of Image colors and Sol...Online Multi-Person Tracking Using Variance Magnitude of Image colors and Sol...
Online Multi-Person Tracking Using Variance Magnitude of Image colors and Sol...
Pourya Jafarzadeh
Ā 

Similar to Similarity Features, and their Role in Concept Alignment Learning (20)

Tdm probabilistic models (part 2)
Tdm probabilistic  models (part  2)Tdm probabilistic  models (part  2)
Tdm probabilistic models (part 2)
Ā 
Genetic Programming for Generating Prototypes in Classification Problems
Genetic Programming for Generating Prototypes in Classification ProblemsGenetic Programming for Generating Prototypes in Classification Problems
Genetic Programming for Generating Prototypes in Classification Problems
Ā 
Clustering techniques final
Clustering techniques finalClustering techniques final
Clustering techniques final
Ā 
Machine Learning and Artificial Neural Networks.ppt
Machine Learning and Artificial Neural Networks.pptMachine Learning and Artificial Neural Networks.ppt
Machine Learning and Artificial Neural Networks.ppt
Ā 
nnml.ppt
nnml.pptnnml.ppt
nnml.ppt
Ā 
call for papers, research paper publishing, where to publish research paper, ...
call for papers, research paper publishing, where to publish research paper, ...call for papers, research paper publishing, where to publish research paper, ...
call for papers, research paper publishing, where to publish research paper, ...
Ā 
Extracting biclusters of similar values with Triadic Concept Analysis
Extracting biclusters of similar values with Triadic Concept AnalysisExtracting biclusters of similar values with Triadic Concept Analysis
Extracting biclusters of similar values with Triadic Concept Analysis
Ā 
EiB Seminar from Esteban Vegas, Ph.D.
EiB Seminar from Esteban Vegas, Ph.D. EiB Seminar from Esteban Vegas, Ph.D.
EiB Seminar from Esteban Vegas, Ph.D.
Ā 
PaperReview_ ā€œFew-shot Graph Classification with Contrastive Loss and Meta-cl...
PaperReview_ ā€œFew-shot Graph Classification with Contrastive Loss and Meta-cl...PaperReview_ ā€œFew-shot Graph Classification with Contrastive Loss and Meta-cl...
PaperReview_ ā€œFew-shot Graph Classification with Contrastive Loss and Meta-cl...
Ā 
Clustering
ClusteringClustering
Clustering
Ā 
TOPIC EXTRACTION OF CRAWLED DOCUMENTS COLLECTION USING CORRELATED TOPIC MODEL...
TOPIC EXTRACTION OF CRAWLED DOCUMENTS COLLECTION USING CORRELATED TOPIC MODEL...TOPIC EXTRACTION OF CRAWLED DOCUMENTS COLLECTION USING CORRELATED TOPIC MODEL...
TOPIC EXTRACTION OF CRAWLED DOCUMENTS COLLECTION USING CORRELATED TOPIC MODEL...
Ā 
An improved teaching learning
An improved teaching learningAn improved teaching learning
An improved teaching learning
Ā 
Text categorization as a graph
Text categorization as a graphText categorization as a graph
Text categorization as a graph
Ā 
Text categorization as a graph
Text categorization as a graphText categorization as a graph
Text categorization as a graph
Ā 
Graph classification problem.pptx
Graph classification problem.pptxGraph classification problem.pptx
Graph classification problem.pptx
Ā 
Text categorization as a graph
Text categorization as a graphText categorization as a graph
Text categorization as a graph
Ā 
Text categorization
Text categorization Text categorization
Text categorization
Ā 
Text categorization as a graph
Text categorization as a graph Text categorization as a graph
Text categorization as a graph
Ā 
Text categorization as graph
Text categorization as graphText categorization as graph
Text categorization as graph
Ā 
Online Multi-Person Tracking Using Variance Magnitude of Image colors and Sol...
Online Multi-Person Tracking Using Variance Magnitude of Image colors and Sol...Online Multi-Person Tracking Using Variance Magnitude of Image colors and Sol...
Online Multi-Person Tracking Using Variance Magnitude of Image colors and Sol...
Ā 

More from Shenghui Wang

Ariadne's Thread -- Exploring a world of networked information built from fre...
Ariadne's Thread -- Exploring a world of networked information built from fre...Ariadne's Thread -- Exploring a world of networked information built from fre...
Ariadne's Thread -- Exploring a world of networked information built from fre...
Shenghui Wang
Ā 
Learning Concept Mappings from Instance Similarity
Learning Concept Mappings from Instance SimilarityLearning Concept Mappings from Instance Similarity
Learning Concept Mappings from Instance Similarity
Shenghui Wang
Ā 
Measuring the dynamic bi-directional influence between content and social ne...
Measuring the dynamic bi-directional influence between content and social ne...Measuring the dynamic bi-directional influence between content and social ne...
Measuring the dynamic bi-directional influence between content and social ne...
Shenghui Wang
Ā 
What is concept dirft and how to measure it?
What is concept dirft and how to measure it?What is concept dirft and how to measure it?
What is concept dirft and how to measure it?
Shenghui Wang
Ā 
Study concept drift in political ontologies
Study concept drift in political ontologiesStudy concept drift in political ontologies
Study concept drift in political ontologies
Shenghui Wang
Ā 

More from Shenghui Wang (13)

Non-parametric Subject Prediction
Non-parametric Subject PredictionNon-parametric Subject Prediction
Non-parametric Subject Prediction
Ā 
Our journey with semantic embedding
Our journey with semantic embeddingOur journey with semantic embedding
Our journey with semantic embedding
Ā 
Linking entities via semantic indexing
Linking entities via semantic indexingLinking entities via semantic indexing
Linking entities via semantic indexing
Ā 
Semantic indexing for KOS
Semantic indexing for KOSSemantic indexing for KOS
Semantic indexing for KOS
Ā 
Contextualization of topics - browsing through terms, authors, journals and c...
Contextualization of topics - browsing through terms, authors, journals and c...Contextualization of topics - browsing through terms, authors, journals and c...
Contextualization of topics - browsing through terms, authors, journals and c...
Ā 
Exploring a world of networked information built from free-text metadata
Exploring a world of networked information built from free-text metadataExploring a world of networked information built from free-text metadata
Exploring a world of networked information built from free-text metadata
Ā 
Ariadne's Thread -- Exploring a world of networked information built from fre...
Ariadne's Thread -- Exploring a world of networked information built from fre...Ariadne's Thread -- Exploring a world of networked information built from fre...
Ariadne's Thread -- Exploring a world of networked information built from fre...
Ā 
Learning Concept Mappings from Instance Similarity
Learning Concept Mappings from Instance SimilarityLearning Concept Mappings from Instance Similarity
Learning Concept Mappings from Instance Similarity
Ā 
Measuring the dynamic bi-directional influence between content and social ne...
Measuring the dynamic bi-directional influence between content and social ne...Measuring the dynamic bi-directional influence between content and social ne...
Measuring the dynamic bi-directional influence between content and social ne...
Ā 
What is concept dirft and how to measure it?
What is concept dirft and how to measure it?What is concept dirft and how to measure it?
What is concept dirft and how to measure it?
Ā 
ICA Slides
ICA SlidesICA Slides
ICA Slides
Ā 
ECCS 2010
ECCS 2010ECCS 2010
ECCS 2010
Ā 
Study concept drift in political ontologies
Study concept drift in political ontologiesStudy concept drift in political ontologies
Study concept drift in political ontologies
Ā 

Recently uploaded

Jual Obat Aborsi Hongkong ( Asli No.1 ) 085657271886 Obat Penggugur Kandungan...
Jual Obat Aborsi Hongkong ( Asli No.1 ) 085657271886 Obat Penggugur Kandungan...Jual Obat Aborsi Hongkong ( Asli No.1 ) 085657271886 Obat Penggugur Kandungan...
Jual Obat Aborsi Hongkong ( Asli No.1 ) 085657271886 Obat Penggugur Kandungan...
ZurliaSoop
Ā 
Spellings Wk 3 English CAPS CARES Please Practise
Spellings Wk 3 English CAPS CARES Please PractiseSpellings Wk 3 English CAPS CARES Please Practise
Spellings Wk 3 English CAPS CARES Please Practise
AnaAcapella
Ā 

Recently uploaded (20)

Interdisciplinary_Insights_Data_Collection_Methods.pptx
Interdisciplinary_Insights_Data_Collection_Methods.pptxInterdisciplinary_Insights_Data_Collection_Methods.pptx
Interdisciplinary_Insights_Data_Collection_Methods.pptx
Ā 
UGC NET Paper 1 Mathematical Reasoning & Aptitude.pdf
UGC NET Paper 1 Mathematical Reasoning & Aptitude.pdfUGC NET Paper 1 Mathematical Reasoning & Aptitude.pdf
UGC NET Paper 1 Mathematical Reasoning & Aptitude.pdf
Ā 
Understanding Accommodations and Modifications
Understanding  Accommodations and ModificationsUnderstanding  Accommodations and Modifications
Understanding Accommodations and Modifications
Ā 
Single or Multiple melodic lines structure
Single or Multiple melodic lines structureSingle or Multiple melodic lines structure
Single or Multiple melodic lines structure
Ā 
SOC 101 Demonstration of Learning Presentation
SOC 101 Demonstration of Learning PresentationSOC 101 Demonstration of Learning Presentation
SOC 101 Demonstration of Learning Presentation
Ā 
This PowerPoint helps students to consider the concept of infinity.
This PowerPoint helps students to consider the concept of infinity.This PowerPoint helps students to consider the concept of infinity.
This PowerPoint helps students to consider the concept of infinity.
Ā 
Unit 3 Emotional Intelligence and Spiritual Intelligence.pdf
Unit 3 Emotional Intelligence and Spiritual Intelligence.pdfUnit 3 Emotional Intelligence and Spiritual Intelligence.pdf
Unit 3 Emotional Intelligence and Spiritual Intelligence.pdf
Ā 
Jual Obat Aborsi Hongkong ( Asli No.1 ) 085657271886 Obat Penggugur Kandungan...
Jual Obat Aborsi Hongkong ( Asli No.1 ) 085657271886 Obat Penggugur Kandungan...Jual Obat Aborsi Hongkong ( Asli No.1 ) 085657271886 Obat Penggugur Kandungan...
Jual Obat Aborsi Hongkong ( Asli No.1 ) 085657271886 Obat Penggugur Kandungan...
Ā 
Unit-V; Pricing (Pharma Marketing Management).pptx
Unit-V; Pricing (Pharma Marketing Management).pptxUnit-V; Pricing (Pharma Marketing Management).pptx
Unit-V; Pricing (Pharma Marketing Management).pptx
Ā 
Basic Civil Engineering first year Notes- Chapter 4 Building.pptx
Basic Civil Engineering first year Notes- Chapter 4 Building.pptxBasic Civil Engineering first year Notes- Chapter 4 Building.pptx
Basic Civil Engineering first year Notes- Chapter 4 Building.pptx
Ā 
How to Create and Manage Wizard in Odoo 17
How to Create and Manage Wizard in Odoo 17How to Create and Manage Wizard in Odoo 17
How to Create and Manage Wizard in Odoo 17
Ā 
ICT role in 21st century education and it's challenges.
ICT role in 21st century education and it's challenges.ICT role in 21st century education and it's challenges.
ICT role in 21st century education and it's challenges.
Ā 
Fostering Friendships - Enhancing Social Bonds in the Classroom
Fostering Friendships - Enhancing Social Bonds  in the ClassroomFostering Friendships - Enhancing Social Bonds  in the Classroom
Fostering Friendships - Enhancing Social Bonds in the Classroom
Ā 
Sensory_Experience_and_Emotional_Resonance_in_Gabriel_Okaras_The_Piano_and_Th...
Sensory_Experience_and_Emotional_Resonance_in_Gabriel_Okaras_The_Piano_and_Th...Sensory_Experience_and_Emotional_Resonance_in_Gabriel_Okaras_The_Piano_and_Th...
Sensory_Experience_and_Emotional_Resonance_in_Gabriel_Okaras_The_Piano_and_Th...
Ā 
Spellings Wk 3 English CAPS CARES Please Practise
Spellings Wk 3 English CAPS CARES Please PractiseSpellings Wk 3 English CAPS CARES Please Practise
Spellings Wk 3 English CAPS CARES Please Practise
Ā 
On National Teacher Day, meet the 2024-25 Kenan Fellows
On National Teacher Day, meet the 2024-25 Kenan FellowsOn National Teacher Day, meet the 2024-25 Kenan Fellows
On National Teacher Day, meet the 2024-25 Kenan Fellows
Ā 
FSB Advising Checklist - Orientation 2024
FSB Advising Checklist - Orientation 2024FSB Advising Checklist - Orientation 2024
FSB Advising Checklist - Orientation 2024
Ā 
Holdier Curriculum Vitae (April 2024).pdf
Holdier Curriculum Vitae (April 2024).pdfHoldier Curriculum Vitae (April 2024).pdf
Holdier Curriculum Vitae (April 2024).pdf
Ā 
Making communications land - Are they received and understood as intended? we...
Making communications land - Are they received and understood as intended? we...Making communications land - Are they received and understood as intended? we...
Making communications land - Are they received and understood as intended? we...
Ā 
ICT Role in 21st Century Education & its Challenges.pptx
ICT Role in 21st Century Education & its Challenges.pptxICT Role in 21st Century Education & its Challenges.pptx
ICT Role in 21st Century Education & its Challenges.pptx
Ā 

Similarity Features, and their Role in Concept Alignment Learning

  • 1. Introduction Three classiļ¬ers Experiments and results Summary Similarity Features, and their Role in Concept Alignment Learning Shenghui Wang1 Gwenn Englebienne2 Christophe GuĀ“eret 1 Stefan Schlobach1 Antoine Isaac1 Martijn Schut1 1 Vrije Universiteit Amsterdam 2 Universiteit van Amsterdam SEMAPRO 2010 Florence
  • 2. Introduction Three classiļ¬ers Experiments and results Summary Outline 1 Introduction Classiļ¬cation of concept mappings based on instance similarity 2 Three classiļ¬ers Markov Random Field Multi-objective Evolution Strategy Support Vector Machine 3 Experiments and results 4 Summary
  • 3. Introduction Three classiļ¬ers Experiments and results Summary Thesaurus mapping SemanTic Interoperability To access Cultural Heritage (STITCH) through mappings between thesauri Scope of the problem: Big thesauri with tens of thousands of concepts Huge collections (e.g., National Library of the Neterlands: 80km of books in one collection) Heterogeneous (e.g., books, manuscripts, illustrations, etc.) Multi-lingual problem Solving matching problems is one step to the solution of the interoperability problem. e.g., ā€œplankzeilenā€ vs. ā€œsurfsportā€ e.g., ā€œarcheologyā€ vs. ā€œexcavationā€
  • 4. Introduction Three classiļ¬ers Experiments and results Summary Thesaurus mapping SemanTic Interoperability To access Cultural Heritage (STITCH) through mappings between thesauri Scope of the problem: Big thesauri with tens of thousands of concepts Huge collections (e.g., National Library of the Neterlands: 80km of books in one collection) Heterogeneous (e.g., books, manuscripts, illustrations, etc.) Multi-lingual problem Solving matching problems is one step to the solution of the interoperability problem. e.g., ā€œplankzeilenā€ vs. ā€œsurfsportā€ e.g., ā€œarcheologyā€ vs. ā€œexcavationā€
  • 5. Introduction Three classiļ¬ers Experiments and results Summary Thesaurus mapping SemanTic Interoperability To access Cultural Heritage (STITCH) through mappings between thesauri Scope of the problem: Big thesauri with tens of thousands of concepts Huge collections (e.g., National Library of the Neterlands: 80km of books in one collection) Heterogeneous (e.g., books, manuscripts, illustrations, etc.) Multi-lingual problem Solving matching problems is one step to the solution of the interoperability problem. e.g., ā€œplankzeilenā€ vs. ā€œsurfsportā€ e.g., ā€œarcheologyā€ vs. ā€œexcavationā€
  • 6. Introduction Three classiļ¬ers Experiments and results Summary Automatic alignment techniques Lexical labels and textual information of entities Structural structure of the formal deļ¬nitions of entities, position in the hierarchy Extensional statistical information of instances, i.e., objects indexed with entities Background knowledge using a shared conceptual reference to ļ¬nd links indirectly
  • 7. Introduction Three classiļ¬ers Experiments and results Summary Instance-based techniques: common instance based
  • 8. Introduction Three classiļ¬ers Experiments and results Summary Instance-based techniques: common instance based
  • 9. Introduction Three classiļ¬ers Experiments and results Summary Instance-based techniques: common instance based
  • 10. Introduction Three classiļ¬ers Experiments and results Summary Pros and cons Advantages Simple to implement Interesting results Disadvantages Requires suļ¬ƒcient amounts of common instances Only uses part of the available information
  • 11. Introduction Three classiļ¬ers Experiments and results Summary Instance-based techniques: Instance similarity based
  • 12. Introduction Three classiļ¬ers Experiments and results Summary Instance-based techniques: Instance similarity based
  • 13. Introduction Three classiļ¬ers Experiments and results Summary Instance-based techniques: Instance similarity based
  • 14. Introduction Three classiļ¬ers Experiments and results Summary Representing concepts and the similarity between them Instance features Concept features Pair features Cos. dist. Bag of words Bag of words Bag of words Bag of words Bag of words Bag of words Creator Title Publisher ... Creator Title Publisher ... Creator Title Publisher ... ... f1 f2 f3 Concept1Concept2 {{ { { { Creator Title Publisher ... Creator Title Publisher ... Creator Term 1: 4 Term 2: 1 Term 3: 0 ... Title Term 1: 0 Term 2: 3 Term 3: 0 ... Publisher Term 1: 2 Term 2: 1 Term 3: 3 ... Creator Term 1: 2 Term 2: 0 Term 3: 0 ... Title Term 1: 0 Term 2: 4 Term 3: 1 ... Publisher Term 1: 4 Term 2: 1 Term 3: 1 ... Cos. dist. Cos. dist.
  • 15. Introduction Three classiļ¬ers Experiments and results Summary Classiļ¬cation of concept mappings based on instance similarity Classiļ¬cation based on instance similarity Each pair of concepts is treated as a point in a ā€œsimilarity spaceā€ Its position is deļ¬ned by the features of the pair. The features of the pair are the diļ¬€erent measures of similarity between the conceptsā€™ instances. Hypothesis: the label of a point ā€” which represents whether the pair is a positive mapping or negative one ā€” is correlated with the position of this point in this space. With already labelled points and the actual similarity values of concepts involved, it is possible to classify a point, i.e., to give it a right label, based on its location given by the actual similarity values.
  • 16. Introduction Three classiļ¬ers Experiments and results Summary Classiļ¬cation of concept mappings based on instance similarity Research questions How do diļ¬€erent classiļ¬ers perform on this instance-based mapping task? What are the beneļ¬ts of using a machine learning algorithm to determine the importance of features? Are there regularities wrt. the relative importance given to speciļ¬c features for similarity computation? Are these weights related to application data characteristics?
  • 17. Introduction Three classiļ¬ers Experiments and results Summary Three classiļ¬ers used Markov Random Field (MRF) Evolutionary Strategy (ES) Support Vector Machine (SVM)
  • 18. Introduction Three classiļ¬ers Experiments and results Summary Markov Random Field Markov Random Field Let T = { (x(i), y(i)) }N i=1 be the training set x(i) āˆˆ RK , the features y(i) āˆˆ Y = {positive, negative}, the label The conditional probability of a label given the input is modelled as p(y(i) |xi , Īø) = 1 Z(xi , Īø) exp K j=1 Ī»j Ļ†j (y(i) , x(i) ) , (1) where Īø = { Ī»j }K j=1 are the weights associated to the feature functions Ļ† and Z(xi , Īø) is a normalisation constant
  • 19. Introduction Three classiļ¬ers Experiments and results Summary Markov Random Field The classiļ¬er used: Markov Random Field (contā€™) The likelihood of the data set for given model parameters p(T|Īø) is given by: p(T|Īø) = N i=1 p(y(i) |x(i) ) (2) During learning, our objective is to ļ¬nd the most likely values for Īø for the given training data. The decision criterion for assigning a label y(i) to a new pair of concepts i is then simply given by: y(i) = argmax y p(y|x(i) ) (3)
  • 20. Introduction Three classiļ¬ers Experiments and results Summary Multi-objective Evolution Strategy Multi-objective Evolution Strategy Evolutionary strategies (ES) have two characteristic properties: ļ¬rstly, they are used for continuous value optimisation, and, secondly, they are self-adaptive. An ES individual is a direct model of the searched solution, deļ¬ned by Ī› and some evolution strategy parameters: Ī›, Ī£ ā†” Ī»1, . . . , Ī»K , Ļƒ1, . . . , ĻƒK (4) The ļ¬tness function is related to the decision criterion for the ES, which is sign-based: LES i = 1 if K j=1 Ī»i Fij > 0 0 otherwise (5)
  • 21. Introduction Three classiļ¬ers Experiments and results Summary Multi-objective Evolution Strategy Multi-objective Evolution Strategy (contā€™) Maximising the number of positive results and negative results are two opposite goals. f1(Ī› | F, L) = #{Fi | K j=1 Ī»i Fij > 0 āˆ§ Li = 1} (6) f2(Ī› | F, L) = #{Fi | K j=1 Ī»i Fij ā‰¤ 0 āˆ§ Li = 0} (7)
  • 22. Introduction Three classiļ¬ers Experiments and results Summary Multi-objective Evolution Strategy Multi-objective Evolution Strategy (contā€™) Evolution process Recombination: Two parent individuals are combined using diļ¬€erent weighting, producing two new individuals Mutation: One parent individual changes itself into a new child individual Survivor selection: NSGA-II
  • 23. Introduction Three classiļ¬ers Experiments and results Summary Support Vector Machine Support Vector Machine Support Vector Machine (SVM) is used as a maximum margin classiļ¬er whose task consists in ļ¬nding an hyperplane separating the two classes. The objective is to maximise the margin separating the two classes whilst minimizing classiļ¬cation error risk.
  • 24. Introduction Three classiļ¬ers Experiments and results Summary Experiments and results Thesauri to match: GTT (35K) and Brinkman (5K) Instances: 1 million books GTT annotated books: 307K Brinkman annotated books: 490K Dually annotated books: 222K
  • 25. Introduction Three classiļ¬ers Experiments and results Summary Feature slection for similarity calculation Ī»j Feature 1 Lexical 2 Jaccard 3 Date 4 ISBN 5 NBN 6 PPN 7 SelSleutel 8 abstract 9 alternative 10 annotation Ī»j Feature 11 author 12 contributor 13 creator 14 dateCopyrighted 15 description 16 extent 17 hasFormat 18 hasPart 19 identiļ¬er 20 isVersionOf Ī»j Feature 21 issued 22 language 23 mods:edition 24 publisher 25 refNBN 26 relation 27 spatial 28 subject 29 temporal 30 title Table: List of the features
  • 26. Introduction Three classiļ¬ers Experiments and results Summary Quality of learning 0 0.2 0.4 0.6 0.8 1 Precision Recall F-measure MRF 1-30 MRF 3-30 ES SVM 0 0.2 0.4 0.6 0.8 1 Precision Recall F-measure 0 0.2 0.4 0.6 0.8 1 Precision Recall F-measure 0 0.2 0.4 0.6 0.8 1 Precision Recall F-measure Figure: Precision, recall and F-Measure for mappings with a positive label (top) and a negative label (bottom). Error bars indicate one standard deviation over the 10 folds of cross-validation.
  • 27. Introduction Three classiļ¬ers Experiments and results Summary Relative importance of features Which features of our instances are important for mapping? Figure: Mutual information between features and labels
  • 28. Introduction Three classiļ¬ers Experiments and results Summary Relative importance of features ES lambdas are not really conclusive ES lambdas that are most inconclusive correspond to the least informative features
  • 29. Introduction Three classiļ¬ers Experiments and results Summary Relative importance of features Important features in terms of mutual information are associated to large MRF weights
  • 30. Introduction Three classiļ¬ers Experiments and results Summary A more detailed analysis Expected important features: Label similarity (1), instance overlap (2), subject (28), etc. Expected unimportant features: Size of the book (16), format description (17) and language (22), etc. Surprisingly important features: Date (14) Surprisingly unimportant features: Description (15) and abstract (8)
  • 31. Introduction Three classiļ¬ers Experiments and results Summary A more detailed analysis Expected important features: Label similarity (1), instance overlap (2), subject (28), etc. Expected unimportant features: Size of the book (16), format description (17) and language (22), etc. Surprisingly important features: Date (14) Surprisingly unimportant features: Description (15) and abstract (8)
  • 32. Introduction Three classiļ¬ers Experiments and results Summary Summary We tried three machine learning classiļ¬ers on the instance-based mapping task, among which MRF and ES can automatically identify meaningful features. The MRF and the ES, result in a performance in the neighbourhood of 90%, showing the validity of the approach. Our analysis suggests that when many diļ¬€erent description features interact, there is no systematic correlation between what a learning method could ļ¬nd and what an application expert may anticipate.
  • 33. Introduction Three classiļ¬ers Experiments and results Summary Thank you