The concept of relevance may be expressed by several words – important, useful, pertinent, valuable, of utility, applicable, helpful, appropriate, significant, topical, correspondence, fit, bearing, matching and others. Following various criticism, this concept became controversial and many definitions of relevance have been offered over the years (Saracevic 2017). Since Vickery (1959), there's been a pattern of discussion about relevance manifestations that continues to this day. To stabilize the concept, Saracevic (2017, p. 19, table) states that “relevance is the A of a B existing between a C and a D as determined by E”, where letters may be filled by combination of terms such as measure, correspondence, document, question, user or by other terms in equivalent roles.

Knowledge organization systems aim at taking advantage of the relevance of a document by indexing its subject, so that it can be found by users interested in them. However, a subject often consists of different themes, that can be cited in the index with different priorities or not cited at all. For example, a document about the application of statistics to railways management can be indexed as statistics – management, or as railways – statistics, depending on the indexing policy. Alberto Cheti recommends that a base theme is identified, that is the theme around which a set of other particular themes are developed in the document, some of which may be expected to be relevant for users. The base theme should be cited first, followed by the particular themes (Gnoli and Cheti 2013).

Hjørland (2017, sect. 2.4 and 2.6) describes the two basic approaches to the notion of subject as related to relevance: content-oriented and request-oriented. In the request-oriented approach, that he favours, indexing should privilege some themes that are expected to be of particular interest to users, even if they are not the document base theme; he provides the example of a 1921 paper that in time has become important for a theme different from its original purpose.

While request-oriented indexing looks like a good service to users, in the context of global open exchange of linked data it also is a hindrance to interoperability, as metadata should be used by a variety of institutions and users the needs of which are not known in advance – some may be more interested in experimental design and some else in crop variation. In this situation, the document-oriented approach may provide a standard reference, where themes are cited according to some known common principle.
Faceted classifications as introduced by Ranganathan (Almeida, Martínez-Ávila and Simões 2017; Satija 2017) follow a standard citation order of the facets that compose a subject for the sake of ordering consistency, known as “PMEST” from the acronym of the facet categories Personality, Matter, Energy, Space and Time. So, within each domain, a faceted KOS gives priority to some facets over others, according to some choice that may be questioned.

A more flexible syntax is offered by the Integrative Levels Classification (ILC), a developing faceted classification system that lists classes of phenomena rather than disciplines (https://www.iskoi.org/ilc/). As a faceted classification, ILC also has a standard citation order of categories, expressed by inverted digits: 9 quality, 8 quantity, 7 part, 6 property, 5 transformation, 4 opposition, 3 agent, 2 place, 1 time, 0 perspective. However, in ILC facets can also be promoted to earlier places to express document-oriented relevance. For example, in case the property facet is more prominent in a document as compared to the part facet, it can be expressed before it in the faceted classmark. This will affect the order in which documents are presented in a browsable list or in query results and their grouping with documents that share the same facet.

Application of ILC facets is demonstrated in Traditional Europe (TradEU), a selected collection of YouTube videos on traditional feasts and other customs of Europe that can be freely tested at https://www.iskoi.org/ilc/tradeu/ (Gnoli et al. 2023). The user can select some facets and launch a search, obtaining a list of matching videos sorted by their faceted classmarks.

For example, some videos of Portuguese traditional events include xn “dances” as a facet. By selecting “dances” from the menu, a query for-xn- is launched and compounds are retrieved which include “dances” as a facet cited in any position. In the video “Entrudo chocalheiro”, dances are the last cited facet, following dressing up and instrument use, which are considered to be more relevant. Sometimes, on the other hand, 50xn “through dances” is cited before 53nrw “with accordions” (or other instruments) despite the default inverted citation order that would prescribe facet 53 to be cited before facet 50: it is so because in those documents indexers have considered dances to be more relevant than the instruments. Similarly, in “Festa dos tabuleiros”, 76w “with flowers” has been promoted before “religious symbol” and “cattle” which would occur before it in the standard citation order, because flowers are obviously the most perspicuous element that can be seen in that feast. Indeed, as it can be appreciated in the video, the tray with flowers and bread is the distinctive element of this event, for which it is known in the rest of the country (Tomar 2023).

On the side of relevance for users, Gnoli and Cheti (2013) recommend that the classmarks where a searched item is the base theme hence occurs at the beginning, such as xn “dance” in the video “Pauliteros de Miranda”, should be shown first, because they are more relevant to the specific search. This can be obtained by Gnoli and Cheti’s double-query method,  although not implemented yet in this prototype.

Its freely-faceted nature thus enables ILC to order search results according to both citation order and a specific searched facet, demonstrating its potential to conceal the representation of relevance within the document with the specific needs and distinctive features of users.