In this work we introduce Coreon Multilingual Knowledge System, a visual tool for concept-based data modeling and curation. Coreon environment is built on the visual paradigm; its goal is to ofer a comprehensible and user-friendly solution for domain experts, well-versed in concept modelling dialects, as well as ad-hoc users, who are new to curation of semantic content. We describe the mechanism of the tool, which, being machine-readable, is powered by the language-agnostic knowledge graph, capably embedding the non-deterministic phenomena of the human language.
Coreon Multilingual Knowledge System (MKS)1 is a visual web-based tool that aims to commoditize the curation and maintenance of structured data[4]. Having witnessed domain experts’ frustration with cumbersome software, our goal is to reduce instrumental barriers and improve working experience for maintainers, who may struggle with obscure data-modeling tools, standards, format intricacies and their limitations[5].
Unlike scrolling through endless flat lists of entries, a dynamic and context-illustrating navigation allows maintainers to scale up and down within the data resource, zooming in on specific parts or branches that require their attention.
With the visual paradigm and model adaptability as its core features, Coreon MKS ofers the user the following key functionalities: • a collaborative environment for easy data curation, accessible for domain experts who had no previous experience with ontology-editing or knowledge-management software. In Coreon, users can create, combine, and edit terminologies, ontologies, taxonomies, controlled vocabularies, and graph-based data collections via familiar drag-and-drop actions, shortcutting extensive training and studying of software manuals; • a customizable browser-based interface, suitable for ad-hoc users and domain experts, for internal and external contributors. The system reacts on the user’s role, hiding preconfigured elements as well as selected properties, should they be flagged as visible only for power users, whereas maintainers have the rest of interaction elements at their disposal; • supporting a variety of established formats: SKOS, TBX, RDF in all relevant syntax lfavours (Turtle, N3, JSON-LD, etc.); • validation of data against custom-established constraints and rules; • data transparency – MKS tracks all record changes, tracing the complete evolution of the data which is often crucial for institutions and highly regulated industries; • seamless interaction and smooth integration with third-party systems and services via RESTful Web API or SPARQL2 protocol that relies on the well-pronounced LLOD standards; • adherence to the Linked Data principles, supporting interoperability and re-usability of language resources [6]; • deployed as light-weight web-based SaaS, it needs no software installation, patches, and configuration updates; • Single-Sign-On mechanism enables simple integration into enterprise authentication environments, based on OASIS’ SAML2 standard.
MKS is a a semantic knowledge repository, with a language-independent knowledge graph as its backbone (see Figures 1, 2). Conceptually, the solution brings the best of Knowledge Organization Systems (KOS) and Terminology Management Systems (TMS) together. MKS leverages the lean structure of SKOS3 yet does not neglect natural language phenomena, bringing along the expressiveness of ISO’s TermBase eXchange (TBX)4.
2https://www.w3.org/TR/rdf-sparql-query/ 3https://www.w3.org/2004/02/skos/ 4https://www.tbxinfo.net/
Coreon is based on an open data model: the user does not have to fit rigid types of complex conceptual models but rather gets maximum flexibility in defining and configuring properties to describe the world. The model is therefore schema-less and easily adaptable for any metadata requirements. Data maintainers have diferent data types at their disposal to best define a concept and its terms (see Figure 3). In production environments, Coreon repositories encounter up to 100 custom properties to fully illustrate a domain.
In a repository, all players receive unique identifiers – individual persistent IDs that unambiguously locate any given element, regardless of whether it is a concept, term, property or a concept relation. In essence, two kinds of information, i.e. knowledge via the semantically linked concepts and language via the terms, are modelled and stored separately from each other while being linked through unique concept identifiers.
From the knowledge perspective, the linking in MKS is performed not at the term but at the concept level. This approach makes concept maps completely independent from the terminological information. By linking entries this way, it becomes possible to model knowledge for phenomena that reflect the non-deterministic nature of the human language, such as word sense ambiguity, synonymy, and multilingualism. Each concept can be populated with unlimited descriptive and documentary information, in dozens of languages. Linking per concept also ensures smooth maintenance of relations without additional data clutter: relation edges are independent from labels, terms and their variants, and other metadata.
From the language perspective, Coreon’s concept model captures the following aspects: • multi-directionality: terms in all languages that belong to the same concept are all stored in one record. Instead of maintaining several databases per language, a user can simply
change the desired language with a mouse-click; • unlimited amount of terms: there is no need to maintain links between terms, neither is there one synonym field. Terms become synonyms as they are simply stored within the same concept; there is no limit on the amount of terms/synonyms for a concept; • term autonomy: each term can carry a full set of descriptive metadata information. This means that all terms in all languages can be exhaustively described; diferent terms can become preferred to support diferent contexts, e.g., in one situation the full form European Central Bank, in another – its acronym ECB.
In contrast with TMSs, the power of Coreon knowledge graph gives users control over large amounts of data. Related concepts (e.g. flat screen , LCD screen, and TFT display) are not just listed under F, L, and T letters, which are alphabetically remote. Rather, they are semantically linked as parent-child concepts and rendered in proximity to each other on the concept map.
A demo read-only MKS containing European Union’s multilingual thesaurus (EUROVOC)5 can be accessed and explored at https://www.coreon.com/dashboard/eurovoc.
Coreon users interact with the environment through four main UI components:
• Concept View (Figure 3), displays all the information of one selected concept, in all languages, together with all metadata properties; • Concept Map (Figure 4), displays the inter-concept relations in a tree-like hierarchy for easy and structured navigation and supports multiple inheritance; • Term List, alphabetic, lexical access to all terms; • Search field, to directly search and locate records in the concept map.
Coreon UI renders not only the hierarchical relations (explicitly shown in the concept-map mode in Figure 4) but also associated relations. To avoid visualization challenges triggered by a potential graph complexity, associated relations are clustered per relation type and currently are displayed for one concept at a time.
We believe that Coreon MKS removes instrumental barriers for conceptual data modeling, becoming a helping hand for domain experts, data maintainers, and other mor and less techsavvy stakeholders. MKS ensures that its users can focus on their domain-modeling tasks, without additional burden of format intricacies and extensive help from IT specialists.
Aside from being the tool for visual data maintenance, MKS ensures a smooth cooperation between domain experts and other professionals, interested in the final outcome of the data modeling activities. Being machine-readable, Coreon MKS promotes resource re-usability and integration into solutions that often benefit from injections of the structured data (e.g. development of virtual assistants, optimization of business processes, fine-tuning of machine translation and machine-learning models). using scatterplot matrix navigation, IEEE transactions on Visualization and Computer Graphics 14 (2008) 1539–1148. [3] N. Tang, E. Wu, G. Li, Towards democratizing relational data visualization, in: Proceedings of the 2019 International Conference on Management of Data, 2019, pp. 2025–2030. [4] M. Wetzel, Multilinguale taxonomien mit coreon. wissens- und sprachmanagement in einer lösung, Rechte, Rendite, Ressourcen. Wirtschaftliche Aspekte des Terminologiemanagements 14 (2014) 41–51. [5] W. Ziegler, Metadaten für intelligenten content, Intelligente Information: Schriften zur
Technischen Kommunikation 22 (2017) 51–66. [6] A. Vasilevich, M. Wetzel, Multilingual knowledge systems as linguistic linked open data for european language grid, in: S. Carvalho, R. R. Souza (Eds.), Proceedings of the Workshops and Tutorials held at LDK 2021 co-located with the 3rd Language, Data and Knowledge Conference (LDK 2021), CEUR Workshop Proceedings, CEUR-WS.org, Zaragoza, Spain, 2021, pp. 126–134. URL: http://ceur-ws.org/Vol-3064/02_Alena_Vasilevich.pdf.