Partitioning and local matching learning of large biomedical ontologies

Abstract

Conventional ontology matching systems are not well-tailored to ensure sufficient quality alignments for large ontology matching tasks. In this paper, we propose a local matching learning strategy to align large and complex biomedical ontologies. We define a novel partitioning approach that breakups large ontology alignment task into a set of local sub-matching tasks. We perform a machine learning approach for each local sub-matching task. We build a local machine learning model for each sub-matching task without any user involvement. Each local matching learning model automatically provides adequate matching settings for each local sub-matching task. Our results show that : (i) partitioning approach outperforms existing techniques, (ii) local matching while using a specific machine learning model for each sub-matching task yields to promising results and (iii) the combination between partitioning and machine learning increases the overall result.

The Impact of Imbalanced Training Data on Local Matching Learning of Ontologies

“…We then perform the partitioning of the input ontologies based on the approach of [13]. This partitioning approach is based on the Hierarchical Agglomerative Clustering (HAC) [17] to produce a set of partition-pairs with a sufficient coverage ratio and without producing any isolated partitions.…”

“…The employed partitioning approach to generate local matching tasks is explained in depth in our previous research work [13].…”

“…Element level features consider intrinsic features of entities such as their textual annotations. Element level features refer to well-known similarity measure matchers, which can be classified into four groups: edit-distance, character-based, term-based and subsequence-based [13]. Structure level features consider the ontological neighborhood of entities in order to determine their similarity.…”

“…Structure level features consider the ontological neighborhood of entities in order to determine their similarity. We previously introduced all these features in a previous research work [13]. Feature selection is performed over each local training set ts ij,q in order to build local classifiers.…”

“…We build local classifiers using the Random Forest algorithm. We have selected Random Forest after comparing the efficiency of several machine learning algorithms for ontology matching learning [13].…”