Ochoa, X. & Duval, E. (2009). Automatic evaluation of metadata quality in digital repositories. International Journal on Digital Libraries, 10(2-3), 67-91.
The frequency of manual quality control of metadata is rapidly declining due to the recent developments in automatic metadata creation and interoperability between digital repositories. However, the result is the occasional absence of any form of quality control that can negatively affect the services provided to the repository users. Various quality metrics for metadata were presented and experiments were run to evaluate these metrics. One compared the quality metrics with quality assessment by human reviewers. The second study evaluated the quality metrics’ ability to discriminate key properties of the metadata sets. The third study tested a practical application in which the quality metrics were tested as an automatic low-quality filter. The metric, Textual Information Content, appeared to be a good approximation of the human reviews and is even able to evaluate quality characteristics that human reviewers cannot. The development of quality metrics will facilitate metadata researchers to monitor the quality of a repository as well as track its growth and the events that can affect it without the need for costly human involvement.
The frequency of manual quality control of metadata is rapidly declining due to the recent developments in automatic metadata creation and interoperability between digital repositories. However, the result is the occasional absence of any form of quality control that can negatively affect the services provided to the repository users. Various quality metrics for metadata were presented and experiments were run to evaluate these metrics. One compared the quality metrics with quality assessment by human reviewers. The second study evaluated the quality metrics’ ability to discriminate key properties of the metadata sets. The third study tested a practical application in which the quality metrics were tested as an automatic low-quality filter. The metric, Textual Information Content, appeared to be a good approximation of the human reviews and is even able to evaluate quality characteristics that human reviewers cannot. The development of quality metrics will facilitate metadata researchers to monitor the quality of a repository as well as track its growth and the events that can affect it without the need for costly human involvement.
No comments:
Post a Comment