To give an example of what I mean here, imagine you are a computer learning agent hooked up to the Google NGram API. You come across an unknown word “Montana”. You guess from syntactic context that “Montana” is a geographic region. Now you search for the trigrams “governor of Montana” and “mayor of Montana”. The latter gets zero hits, while the former gets many, so you conclude “Montana” is a state.
To give an example of what I mean here, imagine you are a computer learning agent hooked up to the Google NGram API. You come across an unknown word “Montana”. You guess from syntactic context that “Montana” is a geographic region. Now you search for the trigrams “governor of Montana” and “mayor of Montana”. The latter gets zero hits, while the former gets many, so you conclude “Montana” is a state.