In this study, we use "The Polish Bibliography" by Estreicher, a comprehensive resource that covers all the Polish prints from 16th-19th centuries, to observe temporal patterns in the history of Polish prints. In order to extract metadata from the "Bibliography" (which lacks any structure), we used a number of LLMs, and experimented with different parameters of our setup.

The presentation will zoom in on the chosen aspects of curating a set of 300.000 records of bibliographic metadata on book-form translations into Polish sourced from the National Library of Poland in the framework of the project: The Hidden Collection: Online database of Polish translated literature since 1918. It will demonstrate the value of combining algorithmic methods of data processing with human-led qualitative verification and amelioration of bibliographic metadata for completeness, accuracy and structure, so that they best reflect translation as a complex cultural and social phenomenon. This will be done on the basis of data on the editions made to our set by more than 60 experts: translators, librarians and researchers of translation, languages and cultures who verified the data in terms of completeness and accuracy and ameliorated them on the basis of their expertise, qualitative research and orientation in existing literature. The presentation will engage with the question of whether, and when, qualitatively ameliorating bibliographic data can, indeed, lead to more reliable results – at different levels of detail and granularity, as well as in consideration of the possible biases in representation of translators and translations in bibliographic metadata.

The Lem Knowledge Graph (LKG) project applies Semantic Web and Linked Data technologies to Cultural Heritage with the aim of creating a comprehensive knowledge base devoted to the works of Stanisław Lem. Its primary goal is to document all of Lem’s writings, while distinguishing between different textual versions and recording the global publication history of these works. The project also seeks to compile a detailed inventory of Lem’s surviving correspondence.

The LKG is designed to go beyond standard bibliographic metadata such as publication dates, publishers, or translators. It also aims to capture information about the content of works and letters, including references to real and fictional persons, other cultural artefacts, and scientific ideas. The knowledge base will be maintained by the Jagiellonian University Museum in cooperation with Lem’s heirs and made publicly accessible through a flexible, multilingual interface intended for both scholars and general readers.

Structurally, the project is organised into three interconnected components. LKG.core collects fundamental bibliographic data, LKG.letters focuses on correspondence, and LKG.content is devoted to modelling the content of individual works. Interoperability is a central design principle, and the graph is built using widely recognised standards, with CIDOC-CRM and its extensions providing the ontological foundation.

The project began in early 2025 with the development of LKG.core and the initial cataloguing of correspondence. Although the original plan relied on integrating existing datasets, differences in their accuracy and completeness have necessitated additional manual data collection and digitisation. The scale and complexity of Lem’s oeuvre—marked by extensive revisions, multiple editions, and fluid boundaries between works—require numerous expert decisions. As a result, the construction of the LKG is not merely a technical task but an interpretative process that will shape future engagement with Lem’s literary legacy.