LIBER 2020 - Workshop: LIBER Linked Open Data Working Group – LOD Publication for Libraries

These are the slides from the LIBER 2020 Workshop, LOD Publication for Libraries, as run by LIBER's Linked Open Data Working Group.

The group is working from 2018-2020 to promote and harmonise the linked data publication of libraries. Special interest is placed in making the library linked data semantically interoperable and on keeping abreast with other similar initiatives both within the library sector and without.

This workshop has two parts. First, a trio of presentations showcasing various aspects of library linked open data. Second, the review of an advanced draft version of a set of best practices for releasing linked open library data. The Working Group conducted a survey on library LOD publishing in 2019 and developed best practices based on the results during the first half of 2020. Based on the discussion in the workshop, a finalized version of the best practices will be released in autumn.

Abstracts

Semantic Interoperability between Bibliographic Models

Integration of library data into the Semantic Web (SW) demands a shift in conceptual data models and data format according to the SW principles and standards. New models have been developed for representing bibliographic information according to the new needs and formats. Well-known models are the Functional Requirements for Bibliographic Records (FRBR) and its consolidated version Library Reference Model (LRM), the FRBR Object-Oriented version (FRBRoo), the Resource Description and Access (RDA) which is based on the FRBR/LRM, and the Bibliographic Framework (BIBFRAME). There also exist cultural heritage models, such as the Europeana Data Model (EDM).

Nowadays, several library and cultural heritage linked datasets have been published. The modelling of the data in each one of these projects has been done differently. Some libraries are using the FRBR model, other libraries are using the BIBFRAME model, while others are using locally-developed models or application profiles. The differences extend to the use of element sets too. Even in cases where the same model is used the element sets are different. Therefore, even though linked data technologies are exploited and technical interoperability is established, the understanding of published library linked data is not ensured. This is a Semantic Interoperability (SI) issue. Unambiguous understanding of bibliographic data is a prerequisite for seamless navigation into the bibliographic universe. Moreover, explicit representation of relationships between bibliographic entities may enable the navigability of bibliographic information even more. Semantic interoperability concerns have already been expressed by scholars in the field.

Towards the goal of contributing to the SI of library data, the project “Semantic Interoperability between Bibliographic Models” has been undertaken by the Ionian University Database & Information Systems Group. This project studies the SI of selected models in the library and cultural heritage domains within the goal to identify good practices facilitating the conversion of one model’s instance to another model’s instance. In detail,

• models' constructs are studied to discover similarities and differences between the models.
• mappings are developed in terms of core classes/entities, inherent relationships, and derivative relationships. Taken into account that derivative relationships are common among the bibliographic universe, there is a focus on them in the framework of this project.
• the mappings are assessed using a testbed.

The models’ study is presented along with key findings regarding semantic similarities and differences between the models. The reconciliation of differences has been attempted in mappings between BIBFRAME-EDM, FRBR-BIBFRAME, RDA-BIBFRAME and BIBFRAME-RDA. The mappings are assessed exploiting Gold Datasets.

This presentation demonstrates the results of the assessment regarding the RDA-BIBFRAME and vice-versa mappings. Both assessments were executed using a testbed with two Gold datasets, Gold RDA and Gold BIBFRAME, that were later compared to the datasets produced by the two mappings.

Based on the results of the RDA-BIBFRAME and BIBFRAME-RDA mappings’ assessment and on the findings from the other mappings, the project concludes that, even though there are semantic differences between the models, semantic interoperability is achievable with some loss of semantics in certain cases. Cataloging policies play a crucial role in minimizing the loss of semantics after conversions. Prerequisites and good practices enabling SI are identified and presented.

---

The Challenge of Interoperability between Entity Models – Share-VDE Experience

Share-VDE (Virtual Discovery Environment) is a library-driven initiative which brings together the bibliographic catalogues and authority files of a community of libraries in a shared discovery environment based on linked data.

Share-VDE had initially adopted BIBFRAME as its entity model to be applied to the incoming MARC data of participating libraries that are converted and published in linked data. The publication of the converted data to the Share-VDE discovery platform raised several challenges. On the one hand there was the need to publish the data reflecting the underlying entity model, also hiding its complexities to the end user; on the other hand, the identification of the “original work” (i.e. the intellectual content of a creation in its highest level of abstraction) was not as simple as expected at the early stages of the process. It became clear that a thorough analysis and reconsideration of the entity model was essential for the identification and reconciliation of entities, as well as for the presentation and distribution of the data elaborated by Share-VDE to its discovery platform and to external parties.

The scenario got even more complex when different bibliographic cultures intertwined in the Share-VDE community: the National Libraries of Norway and Finland and The British Library (that joined Share-VDE along with a pool of libraries from the US and Canada) are traditionally oriented to the IFLA-LRM model, therefore the issue of maintaining as much interoperability as possible between different approaches to entity modeling that converge in Share-VDE needed to be faced.

The solution adopted in order to allow for broader compatibility was to re-shape the model and implement a four-layered adaptation of BIBFRAME that comprises four main entities: svde:Opus, svde:Work, svde:Instance, svde:Item (from the highest to the lowest level of abstraction).

As the discussion among the Share-VDE members and in the broader linked data community moved forward, it came to light that the Share-VDE public discovery platform where the linked data of participating libraries is published has to accurately represent the entity model, but also provide intuitive experience, seamless navigation and rich resources to the end users (both library patrons and professional users). As a result, a new, advanced discovery interface is being developed to harness the potential of linked data and deliver ever more wide-ranging and detailed search results.

The new design of Share-VDE is currently available in a protoype version that is not simply decorative, but that reflects the real interaction with the Share-VDE backend infrastructure and aims to overcome the challenge of publishing straightforward and interoperable linked data.

---

CERL’s resources as Linked Open Data

The Consortium of European Research Libraries (CERL) hosts a number of databases on Early Modern book history. In this talk, we will have a look at two of our main databases: The CERL Thesaurus (CT), which collects places, persons, corporations and printers as they appear in imprints and beyond, and which has been available as Linked Open Data since 2012, and the Heritage of the Printed Book database (HPB) with its over eight million bibliographic records, which we are aiming to publish as LOD in the near future, after a successful prototype was presented in 2019.

The main goal of our work is to provide a sustainable, long-term infrastructure for CERL’s data that remains accessible to all of its heterogeneous audience, ranging from traditional bibliographic research to Digital Humanities applications. We will discuss the challenges in integrating modern standards and technologies with a stable technology stack in production, as well as some of the insights from sharing our latest LOD prototype with researchers from CERL’s community.