Despite the great strides in learning semantic representations of text by decoding the fMRI brain activity, there are still many problems that need to be addressed in mapping the language stimuli and associated brain activations, such as cross-view decoding, mapping the brain activations with transformer based representations, distribution of voxels across the networks in different views. In this paper we are addressing the following problems. 1. Are the brain activity representations same for each view or are there modality-specific as well as modality-independent representations of the same concept? 2. Which of the views has a richer source of information to decode the stimuli? 3. Can we predict the two paradigms using a model trained on a third paradigm? 4. How transformer based Language models associate with fMRI brain activities.
Charan
Manish Gupta Sir and Bapi Raju Sir