A Semantic Search Platform for Old Church
Slavonic Texts and a case study on the Nicene – Constantinopolitan CreedMarianna Napolitano, University of Modena and Reggio Emilia; Giovanni Puccetti, Institute
of Science e Technologies of Information “A. Faedo” - CNR
Abstract
[en]
Semantic retrieval based on Language Models embeddings is an established approach
for information retrieval from large scale textual corpora, however, its application
to digital humanities and specifically to ancient languages is still limited. In this
work we develop a tool for the semantic retrieval of Old Church Slavonic and Church
Slavonic texts and we present an investigation, carried out by experts in Slavic Philology
and Slavic Orthodoxy, of the Nicene - Constantinopolitan Creed in these languages.
The main contributions of the work are two: (1) we describe the creation of a reference
dataset for semantic retrieval in (Old) Church Slavonic, addressing key challenges
such as the limited availability of digital corpora, fragmentation of sources, and
character encoding inconsistencies; (2) we present a tool for semantic retrieval in
these languages, developed as part of this work.
Based on feedback collected from experts using the tool to study the Creed, we find
that it reliably supports its semantic analysis, with the following major strengths:
(a) identifying expressions that are semantically related to the query based on context;
(b) providing access to diachronic sources and enabling assessment of their impact
on the translation and transmission of texts, particularly the Creed, which conveys
the core principles of the Christian faith. The experts also identified some limitations,
which will serve as a basis for future work and further improvement, particularly
when querying texts from manuscript corpora. These issues highlight the need to: (a)
enrich the dataset with manuscript sources using OCR and HTR techniques; (b) account
for script variation and historical or regional orthographic differences.
Our findings highlight how the tool proposed provides both a useful infrastructure
for the study of (Old) Church Slavonic as well as a proof of concept for the application
of semantic based information retrieval in ancient literature.
The tool is available on HORTUS, the ITSERR project website:
HORTUS, an interdisciplinary and distributed Research Infrastructure for Religious
Studies
Entity Extraction in Premodern German Administrative Documents: Accuracy and Cost
in the Age of Generative AIIsmail Prada Ziegler, University of Bern
Abstract
[en]
Named Entity Recognition in historical texts faces unique challenges, including non-standardized
language, transcription noise, and limited annotated material. We evaluate two approaches
on two tasks on the Historical Land Records of Basel, a dataset of 829 German-language
documents from the 14th to 18th century. Task 1 involves classic NER, identifying
persons, locations, and organizations, while Task 2 addresses nested entity annotations,
additionally requiring the identification of grammatical structures such as heads,
appositions, and attributes. We compare generative LLM-based tagging (GPT-4o) with
domain-specific fine-tuned sequence taggers using Flair contextual character embeddings,
examining the impact of variable training data. GPT outperforms Flair in few-shot
scenarios, particularly for low-frequency tags, while Flair achieves superior performance
with larger training sets and longer spans. We find that, no matter which system,
a substantial increase in accuracy can be gained by increasing the amount of training
data. We also analyze annotation cost and speed, showing that manually curated datasets
can be generated efficiently and support well-performing models. Our findings demonstrate
that LLMs provide strong initial results, but lighter traditional systems remain competitive
with sufficient training data.
Design Fiction as Method: The Case of Chinook and Immigration, Refugees and Citizenship
CanadaRalph Padilla, Department of Media and Technology Studies, University of Alberta;
Roya Sharifi, Department of Media and Technology Studies, University of Alberta; Eleyan
Sawafta, Department of Political Science, University of Alberta; Yasmeen Abu-Laban,
Department of Political Science, University of Alberta; Geoffrey Rockwell, University
of Alberta
Abstract
[en]
This article presents a methodological case study of Chinook, a tool employed by the
Canadian government’s department known as Immigration, Refugees and Citizenship Canada
. Chinook assists officers with temporary resident visa applications. To examine
this tool, we use design fiction, a research-through-design method that builds speculative
prototypes in order to explore the social, ethical, and political implications of
technological systems and infrastructures. Given the system's ambiguity, fragmented
documentation, and limited public accountability, we adopt speculative reconstruction
as a critical and interpretive approach. This article relied on court affidavits,
Access to Information and Privacy releases, and legal commentary, to develop a design
fiction prototype and accompanying explainer video to visualize Chinook’s “Decision
Maker” module. Through iterative sketching, prototyping, and annotation, the project
translates bureaucratic and legal evidence into accessible interface logics. By anchoring
speculation in the documentary record rather than in the imagination, and by directing
it at an operational bureaucratic system rather than an emerging technology, the article
reframes design fiction as a method of evidentiary reconstruction, revealing how digital
infrastructures encode discretion and standardization within government decision-making.
Where the Metaphor Breaks: A Review of Theses on the Metaphors of Digital-Textual HistoryLing Zi, Beijing Normal University–École Normale Supérieure de Paris
Abstract
[en]
This review assesses Martin Paul Eve’s Theses on the Metaphors of Digital-Textual History (2024), a study that treats the metaphors of digital writing, such as page, whitespace,
window, and library, as historical artifacts to be excavated and read. Organized as
seven “theses”, the book argues that each such metaphor passes from description, through
rupture, to a prescriptive afterlife that constrains how we imagine text. This review
distinguishes two registers in Eve’s approach: an archaeological one that recovers
material histories, most strikingly that of the PDF, and an interpretive one that
reads the politics of computational metaphor, most inventively in its account of software
as a “reader”. Setting the book against earlier work on new-media metaphor, it finds
the archaeological chapters more persuasive, since their claims rest on archival evidence
rather than analogy, and argues that the book’s methodological range exacts a cost
in coherence. It concludes that the work is best read as a set of provocations rather
than as a system.
A Pidgin Philosophy for Data Visualization at Interdisciplinary CrossroadsMarian Dörk, University of Applied Sciences Potsdam
Abstract
[en]
Archivists, designers, and scholars speak a hybrid language when they negotiate how
collections should appear on screen. Words like ‘generous’ or ‘unfolding’ do work
that neither computer science nor cultural theory does on their own, and usually vanish
once the project ends. I call this a pidgin philosophy for data visualization and
examine its methodological value for research and design. Drawing on cross-cultural
aesthetics and critical interface studies, three such terms are documented in this
paper: wanderer, fold, and shimmer. Each is traced to its sources and checked against
interfaces built for cultural collections. Together they form a grammar comprising
basic principles, a phrase book, and a research framework. Treated as research and
design instruments, these words can help mixed teams collaborate without dissolving
epistemic differences.
From Digitization to AI Integration: Accessing Digital Archives with Insights from
I Trust AIJessica Bushey, School of Information, College of Information, Data & Society San
José State University, California, USA
Abstract
[en]
This article considers how artificial intelligence (AI) tools are reshaping access
to digital archives. While AI offers new possibilities for metadata extraction, transcription,
image analysis, and large-scale discovery, its use in archival contexts raises critical
questions about trustworthiness, ethics, and professional capacity. Drawing on archival
and digital humanities scholarship and recent research initiatives, particularly the
I Trust AI Project, the article identifies three central challenges: protecting the trustworthiness
of records, ensuring that AI adoption aligns with ethical frameworks that resist extractive
practices, and overcoming the AI literacy gap within the archival profession. It argues
that archivists must play a central role in shaping AI integration so that access
systems remain accountable to rights, relationships, and community needs.
Crossing Fonds with the Women’s Labour History Project: Iterating Critical Archival
PracticeSara Diamond, ; Karen Knights,
Abstract
[en]
This paper considers the three phases and five decades of the Women’s Labour History
Project (WLHP) archive. It documents a transmediated feminist archival process, which
has its origins in the pre-digital world and makes use of various technologies, and
arrives at Crossing Fonds, a digital humanities initiative. Transmediation represents the process of moving
records from one medium to another in ways that conserve but also offer new opportunities.
The WLHP engaged in critical archival studies from the 1970s to the 1990s, exploring
human and archival memory, undertaking a forensic search for relevant bibliographic
and archival records regarding women, work, and labour organization, and then growing
into a multi-decade reuse of archival footage to reinterpret diverse histories. Decades
later, the first phase led to Phase Two, the activation of Women’s Labour History
Project series as a component of the larger Sara Diamond fonds—a collection within
the Crista Dahl Media Library & Archive (CDMLA) at VIVO Media Arts through digitization
and generous interfaces built to encourage access. Phase Three is Case Study 1 within
the larger Crossing Fonds Omeka S archival collaboration platform. Digital archives provide the opportunity
to develop new analyses prompted by access to documents from institutional and community
archives. Researchers can redescribe, contextualize, and apply critical archival practices
to consider new relationships, including between items and fonds.