Effective taxonomies for database management

摘要

The development of electronic databases allows for the creation of extremely large data sets of safety-related data. Without adequate taxonomies to categorise and order these data, safety managers run the risk of being swamped ('death by data'). In order to function effectively, taxonomies must enable easy and effective retrieval of data. It is generally agreed in the literature that this implies that taxonomic categories within any taxonomy should be mutually exclusive and exhaustive (MEE). Databases should pass tests of 'inter-rater reliability': in other words, the same event should be classified in the same way by different people. This paper shows the result of an inter-rater reliability trial on an existing taxonomy. The results of this trial indicated that the taxonomy was not functioning in an optimal manner and that the main reason for this was that the 'codes' in the taxonomy were not MEE. In conclusion the paper argues that effective taxonomy building will lead to improved reliability in input and retrieval.