Forgotten Words: Benchmarking NeoBERT for Dementia Detection in Low-Resource Conversational Filipino and English Speech
Researchers have developed new methods for detecting dementia using AI, focusing on both linguistic and acoustic features in speech. One study benchmarks NeoBERT for dementia detection in low-resource conversational Filipino and English, finding that bilingual fine-tuning significantly improves performance. Another approach proposes a multimodal deep learning framework that jointly processes speech and transcript information, using HuBERT for acoustic representations and BERT for linguistic ones, with an attention-based fusion mechanism and a mutual information objective to enhance accuracy. AI
IMPACT Advances in multimodal AI for dementia detection could lead to earlier and more accessible cognitive screening.