TY  - JOUR
AU  - Mestre, Antoni
PY  - 2026
DA  - 2026/10/23
TI  - The Medieval Text and the Modern Question: AI for Multilingual Semantic Retrieval across Centuries
JO  - ICCK Transactions on Machine Intelligence
T2  - ICCK Transactions on Machine Intelligence
JF  - ICCK Transactions on Machine Intelligence
VL  - 2
IS  - 4
SP  - 236
EP  - 256
DO  - 10.62762/TMI.2026.633155
UR  - https://www.icck.org/article/abs/TMI.2026.633155
KW  - natural language processing
KW  - language-aware machine intelligence
KW  - multilingual semantic retrieval
KW  - hybrid information retrieval
KW  - medieval literature
AB  - Multilingual access to medieval literature requires retrieval systems to bridge historical and cross-lingual semantic variation. This article introduces HistRAG, a language-aware framework combining lexical retrieval, multilingual dense representations, historical adaptation, hybrid fusion, and cross-encoder reranking. HistRAG separates the effects of document modernization, query adaptation, and retrieval architecture within a unified experimental setting. The framework is evaluated on a heterogeneous collection of Medieval Catalan literature using equivalent queries in three contemporary languages. Results show that lexical retrieval is highly sensitive to query language, while multilingual dense representations provide more robust cross-lingual access. Fixed hybrid fusion offers no systematic advantage over dense retrieval, and historical adaptations produce heterogeneous, non-additive effects. Cross-encoder reranking achieves the strongest average effectiveness, although its contribution varies across languages and evaluation subsets. These findings support an architecture- and language-sensitive approach to semantic retrieval from historically distant collections and provide a reproducible basis for extending AI-enabled access to other medieval textual traditions.
SN  - 3068-7403
PB  - Institute of Central Computation and Knowledge
LA  - English
ER  - 

@article{Mestre2026The,
  author = {Antoni Mestre},
  title = {The Medieval Text and the Modern Question: AI for Multilingual Semantic Retrieval across Centuries},
  journal = {ICCK Transactions on Machine Intelligence},
  year = {2026},
  volume = {2},
  number = {4},
  pages = {236-256},
  doi = {10.62762/TMI.2026.633155},
  url = {https://www.icck.org/article/abs/TMI.2026.633155},
  abstract = {Multilingual access to medieval literature requires retrieval systems to bridge historical and cross-lingual semantic variation. This article introduces HistRAG, a language-aware framework combining lexical retrieval, multilingual dense representations, historical adaptation, hybrid fusion, and cross-encoder reranking. HistRAG separates the effects of document modernization, query adaptation, and retrieval architecture within a unified experimental setting. The framework is evaluated on a heterogeneous collection of Medieval Catalan literature using equivalent queries in three contemporary languages. Results show that lexical retrieval is highly sensitive to query language, while multilingual dense representations provide more robust cross-lingual access. Fixed hybrid fusion offers no systematic advantage over dense retrieval, and historical adaptations produce heterogeneous, non-additive effects. Cross-encoder reranking achieves the strongest average effectiveness, although its contribution varies across languages and evaluation subsets. These findings support an architecture- and language-sensitive approach to semantic retrieval from historically distant collections and provide a reproducible basis for extending AI-enabled access to other medieval textual traditions.},
  keywords = {natural language processing, language-aware machine intelligence, multilingual semantic retrieval, hybrid information retrieval, medieval literature},
  issn = {3068-7403},
  publisher = {Institute of Central Computation and Knowledge}
}