Port of the deep-fMRI-dataset from [here](https://github.com/HuthLab/deep-fMRI-dataset).

Follows preprocessing from [here](https://github.com/HuthLab/deep-fMRI-dataset).

Provides the (normalized) response to 250 well-predicted voxels in fMRI from one subject (`UTS03`) to natural-language stories.

`text` field contains the 30 words preceding the recording timestep. Rows are from time series data so nearby rows are very related.