The Semantic Bottleneck: Leveraging Semantic Representations for Non-Invasive Speech Decoding
The paper argues that MEG speech decoding works better when it targets sentence meaning first, rather than words or phonemes.
Brain2Semantics2Text maps sentence-level MEG signals into a semantic embedding space, then turns those embeddings back into natural language. The authors frame this as a “semantic bottleneck” that avoids word-level alignment while preserving high-level meaning. They report improved sentence-level results compared with earlier non-invasive Brain2Text methods. ArXiv · AI/CL/LG's note
Brain2Semantics2Text maps sentence-level MEG signals into a semantic embedding space, then turns those embeddings back into natural language. The authors frame this as a “semantic bottleneck” that avoids word-level alignment while preserving high-level meaning. They report improved sentence-level results compared with earlier non-invasive Brain2Text methods. ArXiv · AI/CL/LG's note
score 5