Semantic reconstruction of continuous language from MEG signals

TL;DR

This paper presents a novel data-driven method for decoding continuous language semantics from non-invasive MEG signals, achieving effective reconstruction of text with high similarity to the original speech.

Contribution

It introduces a multi-subject contrastive learning model and a beam search decoding approach for semantic reconstruction from MEG data, advancing non-invasive neural decoding techniques.

Findings

Decoded text showed an average BERTScore of 0.816.

The model effectively leverages subject-specific and shared information.

Results are comparable to previous fMRI-based studies.

Abstract

Decoding language from neural signals holds considerable theoretical and practical importance. Previous research has indicated the feasibility of decoding text or speech from invasive neural signals. However, when using non-invasive neural signals, significant challenges are encountered due to their low quality. In this study, we proposed a data-driven approach for decoding semantic of language from Magnetoencephalography (MEG) signals recorded while subjects were listening to continuous speech. First, a multi-subject decoding model was trained using contrastive learning to reconstruct continuous word embeddings from MEG data. Subsequently, a beam search algorithm was adopted to generate text sequences based on the reconstructed word embeddings. Given a candidate sentence in the beam, a language model was used to predict the subsequent words. The word embeddings of the subsequent words were correlated with the reconstructed word embedding. These correlations were then used as a measure of the probability for the next word. The results showed that the proposed continuous word embedding model can effectively leverage both subject-specific and subject-shared information. Additionally, the decoded text exhibited significant similarity to the target text, with an average BERTScore of 0.816, a score comparable to that in the previous fMRI study.