For a music enthusiast, or even for a librarian, the task of finding all resources connected to Aram Il'yich Khachaturian may be a daunting challenge. Ethnic Armenian of pre-Soviet birth, we find documents in Armenian (“Արամ Խաչատրյան”) and Cyrillic script (“Ара́м Ильи́ч Хачатуря́н”). His audience in the Western Hemisphere is juggling several transliterations, such as “Khachaturian”, “Kaciaturian”, “Chatschaturjan”, Chačaturjan, or “Xač'atryan”, to name but a few.
The fact that these strings all denote the same resource, the epoch-making composer, conductor, and mentor of a generation of Soviet musicians, remains in-actionable. Once entities become things (or resources) that can be linked to each other as a knowledge-base, to rules that can be used for more precise entity extraction, or as automatic quality checks, humans and machines can be made aware of such implicit knowledge. By making facts explicit like: ‘Khachaturian is an Armenian, born in Tiflis, June 6th, 1903, composer of the well-known Waltz from the Masquerade suite (1944)’, one can quickly find resources when searching for ‘Khatchaturian’, even if it doesn’t mention ‘Khatchaturian’ explicitly, and regardless in which language or script the resource is written.
With a semantic layer on top of content, data become meaningful because they are then put in a richer context, and they become actionable because they provide no longer just strings but machine-addressable and -processable things or entities.
Not only when juggling dozens of synonymous denominations for a person, maintaining lists is an everyday-task of information managers. Flat lists often are preferred just in order to avoid the additional overhead often imposed by taxonomies or thesauri.
However, such lack of structure quickly can become a quality issue. One example is dealing with crowd-sourced annotations. Crowd-tagging and folksonomies have become increasingly popular along the Web 2.0 wave. Picklists spawning the entire screen, and separating keyword-spam from keyword-ham are the reverse side of the medal. When data should drive business, actionable knowledge is once again required. Businesses cannot afford not to know if a person entering “coffee” into the search-box is looking for a package of ground beans, or for a place to relax.
A first step towards expressive data is the introduction of simple ontologies, which come at relatively low overhead. When using the Simple Knowledge Organisation Standard (SKOS) from the W3C set of semantic web standards, at the same effort as zoo-keeping a couple of spreadsheets we can get a fully-blown knowledge graph. This is offering concepts uniquely accessible by Universal Resource Identifiers (URIs), hierarchical and non-hierarchical relations, methods that allow for data quality assurance, and and Linked Data networking at the same time.
