Search papers, labs, and topics across Lattice.
This paper introduces the first Slovene ESG sentiment dataset derived from news articles, created using LLM-assisted filtering and human annotation. They benchmarked monolingual (SloBERTa), multilingual (XLM-R), embedding-based classifiers (TabPFN), hierarchical ensembles, and LLMs for ESG sentiment detection. Results indicate that LLMs perform best on Environmental and Social aspects, while fine-tuned SloBERTa excels in Governance classification, demonstrating the feasibility of automated ESG analysis in low-resource languages.
Forget English: LLMs can now reliably extract ESG sentiment from Slovene news, opening doors for automated analysis in overlooked markets.
Environmental, Social, and Governance (ESG) considerations are increasingly integral to assessing corporate performance, reputation, and long-term sustainability. Yet, reliable ESG ratings remain limited for smaller companies and emerging markets. We introduce the first publicly available Slovene ESG sentiment dataset and a suite of models for automatic ESG sentiment detection. The dataset, derived from the MaCoCu Slovene news collection, combines large language model (LLM)-assisted filtering with human annotation of company-related ESG content. We evaluate the performance of monolingual (SloBERTa) and multilingual (XLM-R) models, embedding-based classifiers (TabPFN), hierarchical ensemble architectures, and large language models. Results show that LLMs achieve the strongest performance on Environmental (Gemma3-27B, F1-macro: 0.61) and Social aspects (gpt-oss 20B, F1-macro: 0.45), while fine-tuned SloBERTa is the best model on Governance classification (F1-macro: 0.54). We then show in a small case study how the best-preforming classifier (gpt-oss) can be applied to investigate ESG aspects for selected companies across a long time frame.
