Comment by CokieMonster

That would be nice, but at the moment the algo is at the mercy of the training dataset. The dataset doesn't distinguish adjectives from the ingredient, so this module doesn't :(