Retrieving Biblical Intertextual References in Karen Blixen's Seven Gothic Tales (arXiv Paper)

Forum topic · 小凯 · 2026-09-30

Summary

A 2026 arXiv paper (2609.35765) by Kovács, Conroy, Hershcovich, and Bjerring-Hansen tackles computational detection of intertextual references in literary texts, using biblical allusions in Karen Blixen's Seven Gothic Tales as a case study. The authors build a benchmark of 189 annotated references and evaluate retrieval against all 31,170 verses of historically plausible Danish Old and New Testament translations. They compare TF-IDF and BM25 with multilingual and Danish sentence encoders, test linguistic normalization, and fine-tune a Danish encoder with hard negatives and five-fold cross-validation. Normalized BM25 is a strong zero-shot baseline (overall R@10 = 0.365), while fine-tuning raises DFM-large's R@10 from 0.265 to 0.508, more than doubling allusion performance from 0.138 to 0.339. Human evaluation shows that purely editorial-commentary metrics underestimate model utility: a literary scholar judged 7 of 30 top-ranked false positives to be meaningful additional references. The paper proposes treating retrieval models as heuristic co-readers that both recover documented citations and generate candidates for expert close reading.

Paper Overview

Field: NLP Authors: András Kovács, Alexander Conroy, Daniel Hershcovich, Jens Bjerring-Hansen Published: 2026-09-28 arXiv: 2609.35765

Summary

Identifying intertextual references is central to literary scholarship, but computationally difficult when source material is transformed through paraphrase, allusion, historical language, and translation. The authors investigate this problem through biblical intertextuality in Karen Blixen's *Seven Gothic Tales*.

Approach

Constructed a benchmark of 189 annotated references drawn from the commentary to a critical edition
Evaluated retrieval against all 31,170 verses of historically plausible Danish Old and New Testament translations
Compared TF-IDF and BM25 with multilingual and Danish sentence encoders
Examined the effect of linguistic normalization
Fine-tuned a Danish encoder using hard negatives and five-fold cross-validation

Results

Normalized BM25 provides a strong zero-shot baseline with overall R@10 of 0.365, retrieving each reference within the top ten verses
Fine-tuning improved DFM-large's overall R@10 from 0.265 to 0.508, and more than doubled allusion performance from 0.138 to 0.339
Evaluating solely against editorial commentary underestimates a model's scholarly utility: a literary scholar judged 7 of 30 top-ranked predictions labeled as false positives to be meaningful additional references

Conclusion

The authors propose treating retrieval models as heuristic co-readers — systems that both recover documented citations and generate candidate references for expert close reading.

---

*Auto-collected on 2026-09-30. Original abstract truncated at source.*

Retrieving Biblical Intertextual References in Karen Blixen's Seven Gothic Tales (arXiv Paper)

Summary

Paper Overview

Summary

Approach

Results

Conclusion

Tags