Svenja GuhrLast week (11–13 Feb), I had the pleasure of welcoming Prof. Michaela Mahlberg from @dhss_fau to the UC Berkeley School of Information
#DigitalHumanities #LiteraryComputing
#DigitalHumanities #LiteraryComputing
She inaugurated our #Bellwether Lecture Series with her talk on “Making Sense of the World through Language and Stories: A Digital Humanities Perspective” followed by a lively discussion on concordances, generative AI, and research responsibility in the #DigitalHumanities.
Building on these topics, we spent the following days drafting a new article and preparing our talk, “Soundful Dickens,” analyzing the fictional soundscape of the #Dickens Novel corpus (DNov). We’ll present it next week at the #DHd2026 Conference in Vienna.
#DigitalHumanities #BritishFiction
#DigitalHumanities #BritishFiction
The visit also provided a good opportunity to involve students. At CARL, the Cultural Analytics Research Lab I have been leading since its founding in January. We did a trial run of the talk and discussed current approaches to using language models for classification tasks in #ComputationalLiteraryStudies. #CLS #CHR #StudentResearchLab #CulturalAnalytics
Her stay concluded with a workshop at the AI Futures Lab as part of the #CulturalAnalytics series, where she introduced the new add-on #Flexicon for the concordance web application #CLiC.
Dear Michaela, thank you so much for traveling to the I School and for these productive and collegial days of exchange. 💻✍️📚
And here is the DOI to our long abstract https://doi.org/10.5281/zenodo.18696289
#DHd2026 @DHdKonferenz
#DHd2026 @DHdKonferenz
Soundful Dickens The Narrative Functions of Sounds in Dickens's Fictional Worlds
This paper explores the computational analysis of sound in English-language literary fiction, building on Guhr's (2026) operationalisation of fictional sound events as sound-word-bearing verbal phrases annotated with loudness levels. Originally developed for German prose, the method is adapted here to 19th-century British fiction, using the Dickens Novel Corpus (DNov) as a case study. Rather than relying exclusively on manual annotation, German-language training texts were automatically translated into English using the DeepL API, preserving XML-based annotation spans. These, combined with a single manually annotated English text, were used to fine-tune a pre-trained English BERT model. The results show a surprisingly strong performance compared to similar adaptations in other genres of the same target language. The paper discusses the benefits of using translated annotations and examines sound-related patterns across Dickens's novels using a scalable reading approach to DNov.