Page History

SHARE-VDE is a research and development initiative driven specifically by the library community to facilitate the implementation of BIBFRAME in libraries. This presentation describes the project components including: conversion of over 100 million bibliographic and authority records from 12 North American institutions to BIBFRAME 2.0; reconciliation of entities within the set of converted data, creating a knowledge base of clusters; enrichment of these reconciled clusters with URIs from external sources; and the publication, supply, and management of authority and bibliographical data in RDF. Technical hurdles will be described, along with the solutions adopted, results, feedback, and evidence received from the international library community.

UC-Davis is one of 3 institutions piloting a new suite of linked data services using out-of-the-box services from WikiBase and entities available from FAST, VIAF, and Wikidata to 1) reconcile names for people, organizations, concepts, places, and events against an index based on entities, returning language-tagged headings and persistent identifiers; and 2) create, share and edit entity descriptions while also allowing for the contribution of additional contextual relationships between entities, beyond those that can be found by mining structured data in bibliographic and authority data via an Editor. The partnership aims to build new tools and best practices for designing linked data environments for production-level library workflows.

FOLIO developers and designers are exploring an approach to metadata management in a LOD environment that focuses on empowering the cataloger, and takes a step away from the raw RDF substrate. The internal architecture of the FOLIO project supports multiple models and approaches to metadata management, with common data models that are heavily inspired by the BIBFRAME entity model. The FOLIO design process focuses on the (librarian) user experience, involving subject matter experts collaborating with user experience specialists to produce functional and beautiful designs.

One outcome of the 2017 European BIBFRAME Workshop is a specification for an ILS that supports BIBFRAME. The specification is intended to be used as a basis for evaluating vendor systems and working with vendors to develop ideal ILS components for a linked data production environment. This presentation covers the development of the specification and next steps.

The Ohio State University Libraries (OSUL) uses MarcEdit to integrate reconciliation work directly into daily cataloging workflows. OSUL hopes this work will allow quick implementation of reconciliation in other systems, by enabling other applications to use MarcEdit's reconciliation library and rules file. While MarcEdit isn't an open source application, the underlying platform code, rules files, and sample Ruby components are. On a broader level, this project and the work of other institutions who are implementing the recommendations of PCC's Linked Data Working Group will inform developers and others about the process and limitations around vocabulary services.

This presentation describes the architecture of a robust set of authority services that serve linked data on demand, with response times sufficient for integration into an end-user application. The presentation will demonstrate on-the-fly linkage between BibLeo, a sandbox catalog prototype, and DBpedia, both indirectly via VIAF authority data for persons, and directly for works, merging related images and data from DBpedia with local catalog data.

In a linked data model, libraries can describe manifestations by pointing to existing work descriptions, without having to duplicate those descriptions, but how will this model accommodate work metadata that will be expanded and improved after initial creation? This presentation will describe a model for caching entities, allowing dynamic updating.

Practical implementation of BIBFRAME requires application profiles to specify how the ontology should be used in practice to create metadata. Cornell University libraries developed application profiles for cataloging audio works using BIBFRAME and related extensions; used the SHACL specification to express the application profile, including nested and other complex interactions; and developed code that allows the SHACL profiles to be used directly as configurations for the VitroLib linked data editor. Expressing application profiles in SHACL makes it possible to share the profiles, providing a single record for the decisions and definitions that have been made for particular types of content, and has the potential for reusing portions of the profiles for other content types and workflows.

What would it be like to explore relationships in our collections through an interface that visualizes those connections? Do novel visualizations help users find things they didn't know they were looking for? Does the familiarity of a world map offer meaningful entry points into a collection? Do we actually want graph data to look like a graph? Harvard Library is creating UIs for exploring moving image materials and geospatial datasets described with BIBFRAME and related ontologies, leveraging relationships between entities in the data (collections, works, people, places, genres, materials) and between the data and external linked open data resources (Geonames, ISNI, FAST) to demonstrate queries that would not be available to end-users in a traditional library discovery environment or possible with string-based metadata. This presentation covers the use of open source javascript libraries to create the visualizations, the process by which the underlying metadata was generated, lessons learned about selecting data sources, challenges of integrating manual reconciliation with automated conversion, and the impact of choosing to reconcile or not reconcile metadata with external entities.

Discovery is rarely a single step process. For many searches, library users still seek context and descriptive cues to help them understand, identify and select from the resources returned by their search, even as free-text indexing and ranking algorithms become increasingly sophisticated and powerful. This presentation discusses work at the University of Illinois at Urbana-Champaign to improve discovery in digitized special collections using three approaches: "knowledge cards" constructed with real-time queries of a local triple store and of external LOD services; sidebar descriptions and links fetched in real-time from external LOD services; and the use of Iconclass LOD services to enable multi-lingual keyword searching. These approaches are demonstrated with the Motley Collection of Theatre and Costume Design, the Portraits of Actors collection, the Kolb-Proust Archive, and Emblematica Online. The presentation also covers initial experiments to manually add unique information from these collections to Wikipedia and to create an annotatable visualization of person co-occurrence data mined from Kolb-Proust and made available for retrieval by person identifiers as JSON-LD.

The next phase of LD4P will enhance the Blacklight open source search engine to harness the power of linked data. Features that will be developed include “knowledge panels” in search results to present contextual information from open web sources; browsing powered by a combination of authority files and linked data; semantic search to allow discovery of related concepts and to provide richer geographic-based browsing; and adding RDFa, or microdata, to item pages so that search engines easily find library items. This presentation will seek feedback from attendees on plans for these Blacklight enhancements.