The neuron is not picking out any particular semantic category (tense, sentiment, named entities, etc.) but instead fires broadly on individual lexical tokens – in effect acting as a general “word‐present” detector that lights up whenever a real word (rather than punctuation or whitespace) appears.
