Quality Metrics for Learning Object Metadata

The quality of the learning objects metadata records stored in a repository is important its operation and interoperability. While several studies have tried to define and measure quality for metadata, a scalable and effective way to assess this quality is not currently available. This work converts the fuzzy quality definitions found in those studies into implementable measures (metrics). Several of these metrics are proposed. They are based on the same quality parameters used for human review of metadata: completeness, accuracy, provenance, conformance to expectations, logical consistency and coherence, timeliness, and accessibility. The information requirements to calculate the proposed metrics are also detailed. Some of these metrics are implemented and tested over to two collection of Learning Object Metadata, one of mainly human generated metadata, the other generated by automated means. Early results suggest that the metrics are indeed sensible to quality features in the metadata. Finally, this work recommends further work to validate and calibrate the proposed metrics.