This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
Introduction Mapping method: classification based on instance similarity Experiments and results Summary
Learning Concept Mappings from InstanceSimilarity
Shenghui Wang1 Gwenn Englebienne2 Stefan Schlobach1
1 Vrije Universiteit Amsterdam
2 Universiteit van Amsterdam
ISWC 2008
Introduction Mapping method: classification based on instance similarity Experiments and results Summary
2 Mapping method: classification based on instance similarityRepresenting concepts and the similarity between themClassification based on instance similarity
3 Experiments and resultsExperiment setupResults
4 Summary
Introduction Mapping method: classification based on instance similarity Experiments and results Summary
Thesaurus mapping
Thesaurus mapping
SemanTic Interoperability To access Cultural Heritage(STITCH) through mappings between thesauri
e.g.“plankzeilen” vs. “surfsport”e.g.“griep” vs. “influenza”
Scope of the problem:
Big thesauri with tens of thousands of conceptsHuge collections (e.g., KB: 80km of books in one collection)Heterogeneous (e.g., books, manuscripts, illustrations, etc.)Multi-lingual problem
Introduction Mapping method: classification based on instance similarity Experiments and results Summary
Instance-based techniques
Instance-based techniques: common instance based
Introduction Mapping method: classification based on instance similarity Experiments and results Summary
Instance-based techniques
Instance-based techniques: common instance based
Introduction Mapping method: classification based on instance similarity Experiments and results Summary
Instance-based techniques
Instance-based techniques: common instance based
Introduction Mapping method: classification based on instance similarity Experiments and results Summary
Instance-based techniques
Pros and cons
Advantages
Simple to implementInteresting results
Disadvantages
Requires sufficient amounts of common instancesOnly uses part of the available information
Introduction Mapping method: classification based on instance similarity Experiments and results Summary
Instance-based techniques
Instance-based techniques: Instance similarity based
Introduction Mapping method: classification based on instance similarity Experiments and results Summary
Instance-based techniques
Instance-based techniques: Instance similarity based
Introduction Mapping method: classification based on instance similarity Experiments and results Summary
Instance-based techniques
Instance-based techniques: Instance similarity based
Introduction Mapping method: classification based on instance similarity Experiments and results Summary
Representing concepts and the similarity between them
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
Con
cept
1C
on
cept
2
{{ { {
{Creator
Title
Publisher
...
Creator
Title
Publisher
...
CreatorTerm 1: 4Term 2: 1Term 3: 0...
TitleTerm 1: 0Term 2: 3Term 3: 0...
PublisherTerm 1: 2Term 2: 1Term 3: 3...
CreatorTerm 1: 2Term 2: 0Term 3: 0...
TitleTerm 1: 0Term 2: 4Term 3: 1...
PublisherTerm 1: 4Term 2: 1Term 3: 1...
Cos. dist.
Cos. dist.
Introduction Mapping method: classification based on instance similarity Experiments and results Summary
Classification based on instance similarity
Classification based on instance similarity
Each pair of concepts is treated as a point in a “similarityspace”
Its position is defined by the features of the pair.The features of the pair are the different measures of similaritybetween the concepts’ instances.
Hypothesis: the label of a point — which represents whetherthe pair is a positive mapping or negative one — is correlatedwith the position of this point in this space.
With already labelled points and the actual similarity values ofconcepts involved, it is possible to classify a point, i.e., to giveit a right label, based on its location given by the actualsimilarity values.
Introduction Mapping method: classification based on instance similarity Experiments and results Summary
Classification based on instance similarity
The classifier used: Markov Random Field
Let T = { (x(i), y (i)) }Ni=1 be the training set
x(i) ∈ RK , the features
y (i) ∈ Y = {positive, negative}, the label
The conditional probability of a label given the input ismodelled as
p(y (i)|xi , θ) =1
Z (xi , θ)exp
(
K∑
j=1
λjφj(y(i), x(i))
)
, (1)
where θ = {λj }Kj=1 are the weights associated to the feature
functions φ and Z (xi , θ) is a normalisation constant
Introduction Mapping method: classification based on instance similarity Experiments and results Summary
Classification based on instance similarity
The classifier used: Markov Random Field (cont’)
The likelihood of the data set for given model parametersp(T |θ) is given by:
p(T |θ) =N
∏
i=1
p(y (i)|x(i)) (2)
During learning, our objective is to find the most likely valuesfor θ for the given training data.
The decision criterion for assigning a label y (i) to a new pairof concepts i is then simply given by:
y (i) = argmaxy
p(y |x(i)) (3)
Introduction Mapping method: classification based on instance similarity Experiments and results Summary
Experiment setup
Experiment setup
Two cases:
mapping GTT and Brinkman used in Koninklijke Bibliotheek(KB) — Homogeneous collectionsmapping GTT/Brinkman and GTAA used in Beeld en Geluid(BG) — Heterogeneous collections
Evaluation
Measures: misclassification rate or error rate10 fold cross-validationtesting on special data sets
Introduction Mapping method: classification based on instance similarity Experiments and results Summary
Results
Experiment I: Feature-similarity based mapping versusexisting methods
Are the benefits from feature-similarity of instances in extensionalmapping significant when compared to existing methods?
Table: Comparison using different datasets (feature selected using mutualinformation)
Introduction Mapping method: classification based on instance similarity Experiments and results Summary
Results
Positive-negative ratios in the training sets
0 100 200 300 400 500 600 700 800 900 10000
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Ratio between positive and negative examples in the training set
Mea
sure
s
error rateF−measure
pos
Precisionpos
Recall pos
0 100 200 300 400 500 600 700 800 900 10000
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
Ratio between positive and negative examples in the training set
Mea
sure
s
error rateF−measure
pos
Precisionpos
Recall pos
(a) testing on 1:1 data (b) testing on 1:1000 data
Figure: The influence of positive-negative ratios — Brinkman vs. GTAA
Introduction Mapping method: classification based on instance similarity Experiments and results Summary
Results
Positive-negative ratios in the training sets (cont’)
In practice, the training data should be chosen so as to contain arepresentative ratio of positive and negative examples, while stillproviding enough material for the classifier to have goodpredictive capacity on both types of examples.
Introduction Mapping method: classification based on instance similarity Experiments and results Summary
Results
Experiment IV: Qualitative use of feature weights
The value of learning results, λj , reflects the importance of thefeature fj in the process of determining similarity (mappings)between concepts.
Introduction Mapping method: classification based on instance similarity Experiments and results Summary
Summary
We use a machine learning method to automatically use thesimilarity between instances to determine mappings betweenconcepts from different thesauri/ontologies.
Enables mappings between thesauri used for veryheterogeneous collectionsDoes not require dually annotated instancesNot limited by the language barrierA contribution to the field of meta-data mapping
In the future
More heterogeneous collectionsSmarter measures of similarity between instance metadataMore similarity dimensions between concepts, e.g., lexical,structural
Introduction Mapping method: classification based on instance similarity Experiments and results Summary