Integrating knowledge graphs and multilingual scholarly corpora for domain-adaptive LLMs in SSH
arXiv:2607.05956v1 Announce Type: new Abstract: The integration of Large Language Models (LLMs) into scientific research workflows, particularly for bibliographic discovery and literature synthesis, raises significant methodological, epistemic and regulatory challenges for the Social Sciences and Humanities (SSH), especially with regard to disciplinary diversity, multilingual access to sources and the evaluation of results. This paper presents an on-going use case developed within the European project LLMs4EU and the ALT-EDIC infrastructure, aimed at adapting foundation models to SSH research
Efforts to adapt large language models for specific cultural and academic contexts are intensifying as their general adoption highlights limitations in non-English or specialized domains.
This initiative addresses the critical need for locally relevant and epistemically sound AI tools, mitigating over-reliance on models trained primarily on Western English-centric data and reflecting European values.
The development contributes to multilingual and domain-specific LLMs, fostering greater academic independence and cultural diversity in AI applications for the social sciences and humanities.
- · European academics and research institutions
- · Multilingual AI developers
- · Cultural heritage organizations
- · Monoculture AI providers
- · Generic LLMs without customization options
Domain-adaptive LLMs for SSH reduce biases and improve accuracy in non-technical and multilingual research.
This could lead to increased European autonomy in AI development and a decline in dependency on US-centric AI infrastructure.
Successful adaptation may set a precedent for other regions or disciplines to develop their own culturally and contextually relevant AI models, diversifying the global AI landscape.
This signal links to a primary source. Continuum Brief monitors and indexes it as part of the live intelligence stream — we do not republish source content.
Read at arXiv cs.AI