MaPLE: A MapReduce Pipeline for Lattice-based Evaluation and Its Application to SNOMED CT

Area of Science:

• Computational Linguistics
• Bioinformatics
• Computer Science

Background:

• Non-lattice fragments in ontologies signal structural anomalies requiring quality assurance.
• Traditional methods for extracting these fragments from large ontological systems are computationally intensive.

Purpose of the Study:

• To present a general MapReduce pipeline, MaPLE, for efficient extraction of non-lattice fragments.
• To demonstrate MaPLE's applicability in large-scale ontology quality assurance.
• To analyze structural changes in ontological evolution using MaPLE.

Main Methods:

• Developed a MapReduce pipeline named MaPLE (MapReduce Pipeline for Lattice-based Evaluation).
• Applied MaPLE to analyze 8 versions of SNOMED CT (2009-2014) on a 30-node Hadoop cloud.
• Systematically extracted non-lattice fragments and analyzed structural changes.

Main Results:

• MaPLE achieved an average total computing time of less than 3 hours per SNOMED CT version.
• Non-lattice fragment extraction became feasible for large ontological hierarchies.
• Change rates around non-lattice pairs were up to 38.6 times higher than background concept nodes.

Conclusions:

• MaPLE dramatically reduces the time required for structural analysis of large ontologies.
• The high rate of change in non-lattice fragments highlights their importance in ontological evolution.
• MaPLE enables systematic tracking of structural changes between ontology versions.