prompt: algorithm and tool for automated ontology merging and alignment natalya fridman noy and mark...

PROMPT: Algorithm and Tool for Automated Ontology Merging and

Alignment

Natalya Fridman Noy and Mark A. Musen

Motivation Overview of PROMPT Related Work Knowledge Model PROMPT Algorithm Protégé-based PROMPT Tool Evaluation Discussion Conclusions

Motivation

A variety of ontologies exist in many domain areas, for the purpose of ontology reuses, merging or alignment is necessary

– Merging: create a single coherent ontology that includes the information from all the sources

– Alignment: make the sources consistent and coherent with one another but kept separately

Manually to merge or align is a laborious and tedious work (DARPA’s High performance Knowledge-Bases project)

Many steps in the process of merging or alignment is possible to be automated.

Overview of PROMPT

A Formalism-independent algorithm for ontology merging and alignment

Automate the process as much as possible Guide the users when it is necessary Suggest possible actions Determine conflicts in the ontologies and propose soluti

ons Based on the Protégé-2000 knowledge-modeling envir

onment Can be applied across various platforms

Related Work

Ontology design Object-oriented programming Heterogeneous databases

Related Work-Ontology Design and Integration

Chapulsky et al. 1997, Scalable Knowledge Composition project

Ontomorph Chimaera based on Ontolingua ontology editor Medical vocabularies

Related Work-Object-Oriented Programming

Subject-Oriented programming (SOP)– Subjects: collections of classes that represent subjective views

of, possibly, the same universe that need to be combined

Relies more heavily on the operational methods associated with classes rather than on declarative relations among classes and slots

Alignment is uncommon in composition of object-oriented hierarchies

Related Work-Heterogeneous Databases

The common theme in the research on heterogeneous databases: bridge the gaps on demand by creating an extra mediation layer– Develop mediators– Define a common data model– Specify a set of matching rules

Usually integrated at the syntactic rather than semantic level

Knowledge Model

Classes Slots Facets Instances

PROMPT Algorithm

Input: two ontologies

Output: one merged ontolgy

Gist of PROMPT

Identify a set of knowledge-base operations for ontology merging or alignmentFor each operation, define:

1. Changes that PROMPT performs automatically

2. New suggestions that PROMPT presents to the user

3. Conflicts that operation may introduce and that the user needs to resolve

Ontology-merging Operations and Conflicts

Example (for merging classes A and B to M)

For each slot S that was attached to A and B in the original ontologies

For each superclass of A and B that has been previously copied into the merged ontology

For each class C in the original ontologies to which A and B preferred

For each class C that was a facet value for A or B and that has not been copied to the merged ontology

For each pair of slots for M that have linguistically similar names For each pair of superclasses and subclasses of M that have lingu

istically similar names Check for redundancy in the parent hierarchy for M

Protégé-based PROMPT Tool

Setting the preferred ontology Maintaining the user’s focus Providing feedback to the user Logging and reapplying the operations

Evaluation

Using Protégé-2000 with PROMPT Using generic Protégé-2000 Using Chimaera

Input: two ontologies :A and B contained totally of 134 class and Slot frames.

A: the ontology for the unified problem-solving method Development language

B: the ontology for the method description language

Quality of PROMPT’S Suggestions

% of suggestions that human experts followed

% of conflict-resolution strategies that human experts followed

% of total knowledge-base operation suggestions

90% 75% 74%

PROMPT versus Generic Protégé

PROMPT Generic Protégé

Contents of the resulting merged ontologies

Can find all the classes that should have been merged

Some minor differences in class hierarchy, slot names and types

Number of explicit KB operations that user has to specify

60 16

PROMPT versus Chimaera

PROMT

Chimaera

Correct suggestions

20%

Discussion

The choice of source ontologies Differences between PROMPT and Chimaera

Conclusions

Be able to perform a large number of merging operations

The quality of result should be evaluated The result for larger ontologies is unknown, ne

eds more test Users may be overwhelmed by too many speci

fic suggestions