Supervised tabular models on linguistic features and embeddings outperform zero-shot LLMs for multilingual cognitive impairment detection from speech transcripts, with language-dependent few-shot gains.
Datasets To evaluate cross-linguistic generalisability and per- formance stability, we ran parallel experiments on three languages: English, Slovene, and Korean
1 Pith paper cite this work. Polarity classification is still indexing.
1
Pith paper citing it
fields
cs.CL 1years
2026 1verdicts
UNVERDICTED 1representative citing papers
citing papers explorer
-
Multilingual Cognitive Impairment Detection in the Era of Foundation Models
Supervised tabular models on linguistic features and embeddings outperform zero-shot LLMs for multilingual cognitive impairment detection from speech transcripts, with language-dependent few-shot gains.