Mapping the semantic void III: Exploring neighbourhoods — LessWrong