Working with data in the humanities, we’ve noticed that the debate on classifications is often focused on the definition of the classification and not so much on what it identifies. A well known example is of course ‘nationality’, but also a (historical) occupation/capacity and even seemingly unproblematic classifications like ‘the nineteenth century’ pose several problems.

Looking at data from an object-oriented perspective, using predefined classifications seems counterintuitive. Objects should define themselves by means of their varying attributes. Nodes and clusters emerge on the basis of correlation between objects.

Nevertheless, we understand the need to be able to identify these clusters in a structured manner without the need to perform sequences of filters. These ‘structured clusters’ should be able to be ordered, analysed and explored. For this reason, we have taken up the challenge to equip nodegoat with a functionality that allows for the definition of these clustered by means of fuzzy filtering settings. We have defined this process as ‘reversed classification’. Although we have merely conceptualised the challenge, and have yet to implement this, we want to share our ideas behind this.

In general, classifications emphasise a convention of value and vocabulary. The direction of a classification is outward, relating to the convention unidirectionally. In effect, the classification is unable to communicate/negotiate with the network it classifies. The reversal of classification opens up the convention by disclosing its parameters. Reversal allows the classification to be scrutinised, reconfigured and re-evaluate the objects it classifies.

Simply put: instead of identifying classifications and assigning these to objects in a dataset (like ‘sculptor’ or ‘German’), a user defines a multi-faceted filter spanning multiple datasets in which they define any number of parameters that are associated with a classification. This will reverse the classifying process as the definition of the classification is identified by the exchange between parameters of the classification and attributes of the object.

For the classification of ‘Artist’ this would allow for an inclusion of persons who yield positive on the filter (‘pupil of person who produced paintings’ AND ‘has had paintings in an exhibition’) OR (‘person who studied at School of Visual Arts’ AND ‘has had video installations in an exhibition’). No external labeling or self-labeling of the classification of ‘Artist’ is needed.

By making use of circumstances depending on time and location – defined by Sub-Objects in nodegoat (see FAQ), a reversed classification is able to accompany variating configurations relating to place and time (like an object). Reversed classifications can be employed to cluster people and organisations who are subject to considerable changes over time and in affiliation like activists/freedom fighters/terrorists. Correspondingly, the location and date of a circumstance can also be reversely classified and retrieved from the configuration of the classification. Objects that match ('artifact excavated at a depth of X meters' AND 'in the region of Susa'), could for example be classified with Achaemenid Empire (depending on X) and use the date configured in that classification as its own. The same goes for locations. The domicile of a person can be assessed using titles a person may have had. A location specific title like 'Gouverneur-generaal’ would be classified as ‘Place of residence: Dutch East Indies’.

This reversal works well for concepts and periodisation. When dealing with concepts bound to change over time and space, a reversed classification can be configured to match 17th century objects differently than objects in the 18th century. To facilitate discussions on the definition of periodisations, instead of retagging objects, only the classification has to be reconfigured to match the latest consensus.

We plan to have a first version up and running in the coming months and will use these new functionalities in the project Mapping Notes and Nodes. We will be able to translate research questions into nodes such as ‘who have lived in 17th century Rome in a range of 5km from each other within a 5 month time frame’. Like any other object or classification, a reversed classification is a node as well and can just as easily be mapped geographically, socially, in time, and over distance...