Exploding the metadata myth

Many information managers and architects, myself included, went into battle in the 1990s over metadata: if content was properly labeled with adequate metadata at the point of creation, so the argument went, less effort would be expended later in the information lifecycle in attempting to retrieve content. In contrast, unlabelled content would only offer up any clues by full-text searches. There was a trap there for some of us: by overly insisting that the metadata created for any information object should be intimately associated with the content itself, an excess of zeal led some forget or ignore that the metadata should nonetheless be considered as a distinct set of information and should be managed separately from the content itself. Metadata came to be considered as properties of information, rather than being merely associated with it. Nuance. The problem is that the semantics of metadata change over time. Take an example with the familiar metadata type, “keyword”: what might be considered as an appropriately clear keyword at the moment of content creation may over time become totally inadequate. It may lack granularity: “environment” becomes too vague and insufficiently discriminate if you produce thousands of texts with this sole metadata descriptor, and you might need to be more specific, “nuclear waste management”, “deforestation”, and so forth. The term may fall out of fashion, to be replaced by something that has become more widely used: I can think how the term “racialism” fell into disuse (and disrepute) in the late 1970s, in favour of the term “racism”: some would argue that they are completely different terms, and in context – at the time (and that is what is significant) they were. However, in more general parlance, they would be used to explain the same concept.