A Feminist Approach to Linked Open Data: Making the Women Film Pioneers Project FAIR Pauline Junginger

Tag layer of the DHSS Meta-Tagging Project. The article text is reproduced from the staged copy named in sourceDesc and is not altered; the tags are held separately in standOff so that overlapping and nested spans are preserved.

 A Feminist Approach to Linked Open Data: Making the Women Film Pioneers Project FAIR 2023 Digital Humanities 2023 (DH2023), Graz 10.5281/zenodo.8108135 lod2023-feminist-wfpp lod2023-feminist-wfpp.txt, the project's staged copy 3b17ad1d18bf52405c54c54a306313860a350b2c cc-by digital-humanities

the paper's own words

the tag layer; a sentence written by the tagger about the paper, always anchored to a quotation but not itself a quotation

 39 spans located by searching the stored text for the quote's exact words, allowing any whitespace between them How can research data in feminist film historiography be modelled as Linked Open Data when categorisation reduces the very plurality and gaps the field exists to record? Reports a reversal: from building a project thesaurus to reusing existing ones, and what that cost in categories. method linked open data for feminist film historiography system-or-tool-buildingsystem-or-tool-building Reusing existing thesauri rather than building one forces the project's categories down to those that carry across datasets, which is what makes the data linkable at all Extraction moves from manual annotation to named entity recognition once the target is linked data rather than a bespoke vocabulary Existing authority files are treated as objects of evaluation rather than given infrastructure, because their fitness for feminist film history is not assumed The paper reports only preliminary results; it is a poster describing a methodological approach rather than a completed, evaluated system.what this limits: everything reported: THIS POSTER OUTLINES THE METHODOLOGICAL APPROACH OF THE PROJECT AND PRESENTS PRELIMINARY RESULTS The scope was narrowed mid-project from nine modelling categories to four, restricting coverage to filmographic data.what this limits: the coverage of the thesaurus: THE NINE CATEGORIES INITIALLY IDENTIFIED ... WERE REDUCED TO FOUR The authors state as an open problem that LOD modelling forces categorisation, standardisation and a reduction of complexity that may not accommodate the plurality of feminist film historiography.what this limits: any LOD representation of the field's plurality: MODELING RESEARCH DATA AS LOD REQUIRES CATEGORIZATION AND STANDARDIZATION, AND THUS A REDUCTION OF COMPLEXITY Modelling as LOD requires standardisation, which reduces exactly the complexity the field is trying to preservewhat this limits: the same scope as limitations[2] - standardisation reduces exactly the complexity the field exists to preserve Classification is taken to be never neutral, so the vocabularies being reused carry values that are not the project's ownwhat this limits: every reused vocabulary: CLASSIFICATION PRACTICES ARE NEVER NEUTRAL BUT ALWAYS ENTANGLED WITH POLITICAL ATTITUDES AND VALUES The gaps left by lost sources and undocumented feminized labour have no obvious representation in a data modelwhat this limits: anything the data model is taken to represent completely: THE NUMEROUS BLIND SPOTS AND KNOWLEDGE GAPS THAT EXIST DUE TO LOST SOURCES AND THE LIMITED DOCUMENTATION OF FEMINIZED LABOR have no obvious representation The poster reports an approach and preliminary results rather than outcomes, so nothing here is yet evaluatedduplicate of: limitations[0] · what this limits: the same scope as limitations[0] - an approach and preliminary results, nothing yet evaluated authority fileauthority file Virtual International Authority FileVirtual International Authority File thesaurus - project-specific: built bottom-up for this resource's own staff and users, which was the original plan thesaurus - existing and reused: an external vocabulary adopted so the data links outward, which is what the project switched to categories - the nine identified for modelling a thesaurus, covering the resource's whole conceptual range categories - the four kept for filmographies, which is what survived the decision to link outward claim_type: methoda second reading reaching the same value. Recorded: method — note: a second reading reaching the same value from a different sentence. Not applied - a person decides. phenomenon: linked open data for feminist film historiographya second reading reaching the same value. Recorded: linked open data for feminist film historiography — note: a second reading reaching the same value from a different sentence. Not applied - a person decides. builds-on → digital humanities work on text as data, digital publishing, categorization practices and metadata (Blaney et al. 2021; Flanders / Jannidis 2018) builds-on → critical classification scholarship (Bowker / Star 1999; Drabinski 2013; D'Ignazio / Klein 2020) extends → existing work modelling film heritage data as Linked Open Data, including Wikidata case studies supports → the position that classification practices are never neutral (Bowker & Star, Drabinski, D'Ignazio & Klein) Linked Open Data —operationalised-as→ FAIR data principlesstrength of the relation: 1 Audit 2026-09-28: DOUBTFUL - read against its span and kept, but LOD serves FAIR interoperability; operationalised-as is the wrong type.Linked Open DataFAIR data principles Wikidata —is-a→ authority datastrength of the relation: 1 Audit 2026-09-28: DOUBTFUL - read against its full span and kept, but the span lists Wikidata among 'thesauri and authority files'; 'authority data' is a variant.Wikidataauthority data Virtual International Authority File —is-a→ authority datastrength of the relation: 1 Audit 2026-09-28: DOUBTFUL - read against its full span and kept, but is-a rests on the word 'Authority' inside the name; the span is a bare list of resources.Virtual International Authority Fileauthority data Named Entity Recognition —contrasted-with→ CATMAdiffers in: THE AXIS IS MANUAL VERSUS AUTOMATIC EXTRACTION. 'whereas the initial concept included a manual text analysis and annotation with CATMA to build the thesaurus in a bottom-up approach, the filmographic data is now extracted using Named Entity Recognition and then modeled as LOD.' A project describing its own method change, so the two are alternatives for one task rather than rival accounts of anything. · strength of the relation: 1Named Entity RecognitionCATMA Named Entity Recognition —operationalised-as→ filmographiesstrength of the relation: 0.9 Audit 2026-09-28: DOUBTFUL - read against its span and kept, but NER is the extraction method for filmographic data; operationalised-as is the wrong type.Named Entity Recognitionfilmographies Virtual International Authority File —is-a→ authority filestrength of the relation: 1 Audit 2026-09-28: DOUBTFUL - read against its full span and kept, but the span lists VIAF among 'thesauri and authority files'; is-a is loose.Virtual International Authority Fileauthority file The project reversed its approach mid-way: reusing existing thesauri beat building its own. The nine categories first identified for modeling the thesaurus were reduced to four. the resource's own use of metadata had been very limited Women Film Pioneers ProjectWomen Film Pioneers Project 9 → 4 — modelling categories before and after the shift to reusing existing thesauri nine categories reduced to four — the concrete cost of switching from a bespoke thesaurus to reusable vocabularies

Introduction

The Women Film Pioneers Project (WFPP) is an established online resource for research on women in early cinema. The project focuses on telling individual stories of women film pioneers and making blind spots visible in order to draw attention to the major role played by women in the early days of film history (Gaines et al. n.d.). Initially conceived in the early 1990s as a classic book project, in 2013 the WFPP was published as a digital platform with three primary goals: to make research findings available, to stimulate further research, and to challenge traditional film historical narratives that fail to mention women's contributions. However, although structured metadata is central to making research data and digital publications visible, accessible, and reusable (Baca 2016; Flanders / Jannidis 2018), WFPP's application of metadata has been very limited. Therefore, my project aims to create structured metadata for the WFPP, enrich it with authority data, and link it in a sustainable way to other datasets on women film pioneers.

Related work

In addition to feminist film historiography, the digital humanities are a central point of reference for this project, particularly approaches that address accessibility of text as data, digital publishing, categorization practices, and metadata (Blaney et al. 2021; Flanders / Jannidis 2018). In order to make metadata interoperable in the sense of the FAIR data principles, it can be made available as Linked Open Data (Schmidt et al. 2022). In the field of film heritage institutions, the modeling of film data as Linked Open Data has been discussed for several years (Heftberger 2019), including case studies with Wikidata (Heftberger / Duchesne 2020).

Methodology

While the initial goal of the project was to develop a thesaurus tailored to the needs of WFPP’s staff and users, it became clear as the project progressed that a more promising approach would be to focus on reusing existing thesauri to make the data FAIR. This conceptual shift from developing a project-specific thesaurus to opening up the collection by sustainably linking disparate datasets on women film pioneers had several implications for the project's methodological design. First, the nine categories initially identified for modeling the thesaurus (person, film work, occupation, place, technique, institution, film genre, theme, and publication) were reduced to four categories relevant to the structured collection of filmographies: person, film work, institution, and place. By creating FAIR metadata using the WFPP filmographies as an example, the project explores how the FAIR principles can be applied in feminist film historiography. Second, whereas the initial concept included a manual text analysis and annotation with CATMA to build the thesaurus in a bottom-up approach, the filmographic data is now extracted using Named Entity Recognition and then modeled as LOD. Third, an important step in the methodology is to analyze and evaluate existing thesauri and authority files such as the Getty Art & Architecture Thesaurus (AAT), Friend of a Friend (FOAF), Virtual International Authority File (VIAF), Gemeinsame Normdatei (GND), and Wikidata regarding their usability for feminist film historiography.

Critical reflection

From a feminist perspective, when working with Linked Open Data, it is important to ask what assumptions, biases, and power relations are inscribed in Semantic Web technologies as well as in authority files and vocabularies. Furthermore, it also involves an examination of the epistemic conditions of a structured data approach and a reflection on the specific requirements of feminist film historiography for applying the FAIR data principles. How can the plurality of feminist film historiography be accounted for when modeling research data as LOD requires categorization and standardization, and thus a reduction of complexity? How can we incorporate feminist values into the design of categories if we assume that classification practices are never neutral but always entangled with political attitudes and values (Bowker / Star 1999; Drabinski 2013; D'Ignazio / Klein 2020)? In the context of feminist film historiography, we also have to ask how the numerous blind spots and knowledge gaps that exist due to lost sources and the limited documentation of "feminized labor" (Hill 2016: 5) can be represented in data-based approaches (Dang 2020).

This poster outlines the methodological approach of the project and presents preliminary results.

Acknowledgements

This project is carried out within the framework of the BMBF-funded research group “Aesthetics of Access. Visualizing Research Data on Women in Film History (DAVIF)”.

Project website: https://www.uni-marburg.de/en/fb09/institutes/media-studies/research/research-projects/davif [01.05.2023].