Decoding silent reading from non-invasive EEG
Summary
A machine-learning study demonstrates that open-vocabulary word-level information can be decoded from EEG during silent reading. The team uses a 19-channel dry-electrode EEG setup and a CLIP-style objective to align EEG windows with word embeddings from a large language model, achieving above-chance word decoding that scales with data volume and persists even with partial electrode removal.