ontology and knowledge representation · BEGINNER LESSON
Alignment
Ontology alignment is the process of comparing two ontologies and finding which concepts in one correspond to concepts in the other with the same intended meaning. It is used to relate different ontologies, often to support comparison or integration.
What it is
In these sources, alignment means ontology alignment: a comparison between two ontologies, where one may be treated as a reference and the other as a candidate. The goal is to find corresponding concepts, often equivalent concepts, across the two ontologies. This helps compare ontology versions or support integration of ontologies from different sources.
[1] THE EVALUATION OF ONTOLOGIES [2] Use artificial neural network to align biological ontologies - PMC
What is measured
Alignment methods measure how well concepts, relations, or structures in one ontology match those in another. Some methods compare concept names, some compare concept properties and relationships, and some use distances or scores to estimate how similar the aligned ontologies are. Alignment can give partial information about quality, but it is not a full evaluation metric.
[1] THE EVALUATION OF ONTOLOGIES [2] Use artificial neural network to align biological ontologies - PMC
How it works
The sources describe several mechanisms. One common approach compares vocabulary: terms that label concepts are matched using lexical similarity, but this is limited by synonymy, where different words mean similar things, and ambiguity, where one word can mean different things. Another approach compares structure, such as positions in a concept lattice, meaning the organized network of concepts around each node. A learning-based approach can use an artificial neural network (ANN), which is a machine-learning model that adjusts weights for different semantic aspects such as names, properties, and relationships. In the biological ontology example, these weights are learned from a subset of concepts and then used to compare others.
[1] THE EVALUATION OF ONTOLOGIES [2] Use artificial neural network to align biological ontologies - PMC
Limitations
Alignment is difficult because the right reference ontology may be hard to choose, and because there is no gold-standard reference ontology for complete evaluation. Term matching can fail when labels are synonyms, ambiguous, or written in complex multi-word forms. Structure-based methods work best when ontologies are very similar, such as different versions of the same ontology. Learning-based methods can need a lot of data and may run slowly, while biological ontologies may be large and contain very few usable instances.
[1] THE EVALUATION OF ONTOLOGIES [2] Use artificial neural network to align biological ontologies - PMC
SEE THE IDEA
How ontology alignment is done
The method starts with ontology O1 and ontology O2, which may describe the same domain in different ways.
Algorithms check labels, concept properties, relationships, and sometimes lattice position.
The method proposes correspondences such as equivalent concepts or related parent-child pairs.
The result gives an estimate of how similar the ontologies are, but not a full gold-standard evaluation.
Input ontologies are compared by labels, structure, or learned weights to produce concept correspondences and a similarity judgment.
Ontology alignment maps concepts across ontologies by comparing semantic clues, but the result is only a partial measure of quality.
TRY IT YOURSELF
Test your understanding
Choose an answer, then check the explanation.
PUT IT TOGETHER
Match each concept to its meaning
Drag a label onto its description, or select a label and then select a description. This also works with a keyboard.
4 pairs to match
Sources and further reading
This general lesson was generated from public source material. It explains a concept; it does not interpret a private paper or provide medical advice.