Gespeichert in:
| 1. Verfasser: | |
|---|---|
| Format: | Recurso digital |
| Sprache: | Deutsch |
| Veröffentlicht: |
Zenodo
2026
|
| Schlagworte: | |
| Online-Zugang: | https://doi.org/10.5281/zenodo.18725927 |
| Tags: |
Tag hinzufügen
Keine Tags, Fügen Sie den ersten Tag hinzu!
|
Inhaltsangabe:
- <p><em>Spurenlese</em> explores a recurring interpretive tension in digital history: what, exactly, are the “results” of digital methods for historians? Many digital humanities tools reliably yield statements about <em>processed data</em> (token streams, similarity scores, extracted entities), while historical scholarship ultimately seeks statements about <em>people, practices, and situations in the past</em>. The step from one to the other often happens too quickly, by habit or epistemic reflex.</p> <p>To keep this gap visible without abandoning computation, the talk proposes the <strong>“trace” (Spur)</strong> as a theory figure. In this framing, algorithmic outputs are not treated as historical findings by default, but as <strong>navigational instruments</strong>: they help select patterns that can then be interpreted as traces of situated human practices (copying, curation, canonization, exclusion, authorship, authorization).</p> <p>The argument is developed through two schoolbook projects at the <strong>Leibniz Institute for Educational Media | Georg Eckert Institute (GEI)</strong>:</p> <ul> <li> <p><strong>SchulbuchEvolution (2024–2027)</strong>: uses text reuse detection (including n-gram shingling and similarity metrics) to identify dense overlaps between volumes in the GEI-Digital corpus. These overlaps function as entry points for reconstructing editorial and curatorial practices and for asking how specific production contexts shaped the genre over time.</p> </li> <li> <p><strong>DigiRel (2024–2027)</strong>: combines OCR with <strong>named entity recognition (NER)</strong> and <strong>named entity linking (NEL)</strong> (including links to the <strong>GND</strong>) to extract and connect actors from title pages and bibliographic paratexts. The resulting structured data can be read as traces of historically shifting regimes of authorship and authorization, and as a basis for network-oriented inquiry, while foregrounding error sources and validation work.</p> </li> </ul> <h3>Contents of this deposit</h3> <ul> <li> <p><strong>Slides (PDF)</strong> for the BBF DHLabs “Last Friday’s Lab Talk” (Berlin, 29 Aug 2025).</p> </li> <li> <p><strong>Talk manuscript (PDF)</strong> corresponding closely to the spoken version. The manuscript retains a talk register (not rewritten into polished prose) and includes a selective bibliography. Some technical explanations (especially n-gram shingling) remain easier to follow alongside the slides.</p> </li> </ul> <h3>Notes on reuse and scope</h3> <p>These materials are shared as talk-level research output: they aim to make an argument legible and cite the intellectual context. They should be read as a contribution to ongoing discussion about how computational models and hermeneutic interpretation can be coupled without collapsing into each other.</p>