My last post discussed the benefits of using hierarchical relationships in the form of a taxonomy or classification scheme to go from the general to the specific. However, it’s useful to explore the other ways in which metadata can be meaningful and contextually useful when it is not part of a taxonomy structure. Or maybe to focus on how metadata can be meaningful by focusing more on the associative and equivalence relationships present in a taxonomy, instead of just the hierarchical ones. Perhaps there is even another type of relationship that has not yet been discussed or defined to go from the general to the specific.
I’ve often heard the expression “metadata is king,” in fact I even wrote about it in a post from last year. However, from an information management perspective, the focus on and use of metadata is only part of the equation. Using metadata to get from the general to the specific is only beneficial when it is set up properly and defined consistently. Another key element is the context needed to interpret and leverage metadata effectively. This is why those relationships are so important, even if they are not hierarchical in nature.
I always think social media is a great macro-level example of metadata going horribly wrong and fantastically right at the same moment. The user-generated metadata, such as tags, #hashtags, keywords, etc. exemplifies the many things that can, and in fact do go wrong, with inconsistently applied terms that have no relationships or meaning with each other. It creates chaos and negates the benefits of using these terms to aid the user in searching for and retrieving similar items. Have you ever tried searching on Twitter? I have and my results are sometimes successful and sometimes not, but always contain a lot of random, off-topic tweets.
Machine-generated metadata, however, can be incredibly effective for generating analytics on large data sets. The backend of each tweet, for example, contains dozens of defined metadata fields, all of which can be used to generate analyses of huge data sets, such as language or travel patterns.
So what is the best way to go from the general to the specific? Perhaps the best solution is to combine the best of both worlds and create a faceted classification structure to leverage the benefits of metadata while still preserving the contextual connections that provide so much meaning and consistency for the items being organized. In this way the metadata terms retain meaning because they have a direct relationship with the object being managed, while still providing options for different types of filtering and browsing to achieve the end result.
Or maybe we have to re-imagine the whole system and build a foundation that is not based on a hierarchy. What are your thoughts?