Looking at the activations, I notice that the neuron has very low activations (all 0 or close to 0, with only one activation of 0.4766 for "some") across all the examples shown. However, examining the content of the documents where it