Arcarium (ARC) will be an information system made up of a knowledge base and supporting interfaces to facilitate a wide range of manuscript research: historical, philological, amateur, etc.

First of all, the knowledge base will accumulate information extracted from the metadata of items catalogued under the DiHeLib Project: documents belonging to the *Sammlung Autographa* Collection. This knowledge base will be constructed using knowledge graph technology. Various units of information – such as names, institutions, geographic locations, document titles, dates, and so on – will be integrated into a structured network linked through diverse relations (e.g., authorship of a document, sender or recipient of a letter, mentioning something, correspondence with someone). Whole graph will be based on a dedicated formal model (The Arcarium Ontology; TAO). Individual documents from the Collection will have a detailed representation in the knowledge base, accounting for both the complete document and its constituent parts, like pages or cards. Each page will be associated with a link to its photocopy hosted on the servers of the Jagiellonian Library.

Given that many of the items described within the Project — including documents, individuals, locations, institutions — already have representations in existing globally accessible knowledge bases, such as Wikidata and Geonames (and other Linked Open Data data sets), the development of ARC will emphasise tight integration with these external resources. The primary focus will be on Kalliope, the unified manuscript and publication catalog managed by the Staatsbibliothek zu Berlin, and Europeana, the pan-European cultural heritage knowledge base.

The designed modules and interfaces to the knowledge base will facilitate convenient information retrieval and browsing of document photocopies. Using the knowledge graph thus constructed, the ARC system will also offer modules, founded on artificial intelligence methods and algorithms, to facilitate specialized research in the field of digital humanities, such as advanced visualization of collected data, reconstruction of document history based on provenance information, reconstruction of networks of connections between individuals, institutions, locations and documents, (semi-)automatic methods of knowledge enhancement based on graph embeddings, automatic handwriting recognition, visual analysis and semantic annotation of non-text documents (drawings, music scores, etc). Moreover, the open modular architecture of ARC will allow for its continued expansion or refinement by the research community. The knowledge base will also accommodate the diverse perspectives of scholars on certain contested matters: it will be possible to model multiple alternative relations between the units constituting the knowledge base such as an authorship and dating of particular documents.