On the way to measure KG transparency: Formalizing transparency - Requirements and first models
Abstract
The demand for transparency is high and increases in many domains including everyday life (government, business, environment, health, personal data management...) and science. Although very often cited and desired, the contours of this notion often remain rather vague and can even differ significantly from one context to another. Such lack of precision facilitates the possibility for an element to be qualified as transparent by its promoters without this element meeting the expectations of outside observers. For example, many governments have made significant efforts to make open data available. However, from the public's point of view, there are limitations that can hinder transparency: file formats that limit queries, data formats that are difficult to use, little information on how certain data are produced, etc. The use of norms and standards for data publication (data format, access and query protocols…) brings an obvious improvement. In particular, knowledge graphs make it possible to represent, publish, make accessible and searchable information in any field. Whatever the considered interpretation of transparency, the use of a public KG constitutes a major element that can only improve transparency. However, using a KG is not in itself sufficient to fully guarantee transparency.
In this report, our first objective is therefore to better understand and formally define the concept of transparency. First, we present existing definitions of transparency, as well as many of its companion concepts that are very often used to complement and to clarify it: accessibility, accountability, provenance, privacy, openness, etc. The synthesis makes us distinguish between notions that characterize data sources and those that characterize observation needs and leads us to pose our own informal definition of transparency. Then, we propose a formal definition of the concept of transparency in a general context which we then adapt to knowledge graphs.
Origin | Files produced by the author(s) |
---|---|
Licence |
Copyright
|