Introducing MEG-MASC a high-quality magneto-encephalography dataset for evaluating natural speech processing

Area of Science:

• Neuroscience
• Cognitive Science
• Linguistics

Background:

• Understanding brain responses to naturalistic speech is crucial.
• Existing datasets may lack the temporal resolution or annotation detail required for advanced analyses.
• Magnetoencephalography (MEG) offers excellent temporal resolution for studying speech processing.

Purpose of the Study:

• To introduce the MEG-MASC dataset, a novel resource for studying brain responses to speech.
• To provide a benchmark for large-scale encoding and decoding analyses of temporally-resolved brain activity.
• To facilitate reproducible research in speech neuroscience.

Main Methods:

• Collected raw magnetoencephalography (MEG) recordings from 27 English speakers.
• Participants listened to naturalistic stories from the Manually Annotated Sub-Corpus (MASC).
• Word and phoneme onsets/offsets were time-stamped and data organized using the Brain Imaging Data Structure (BIDS) format.

Main Results:

• The MEG-MASC dataset contains detailed recordings with precise temporal annotations.
• Validation analyses demonstrated the utility for temporal decoding of phonetic features and word frequency.
• The dataset supports reproducible research through publicly available code and data.

Conclusions:

• The MEG-MASC dataset is a valuable resource for advancing speech processing research using MEG.
• It enables robust encoding and decoding models of neural responses to speech.
• Public availability promotes transparency and collaboration in the field.