Entity may not be picked up exactly by the speech recognition. Instead
of doing an exact string match, look for names which are phonetically
similar and choose the best match above a certain threshold.
This helps as entity names may often be pronoun's (Dave's bedroom) and
this allows for some minor mistakes i.e "the bedroom" will successfully
match against a switch called "bedroom"
It introduces a new library dependency, fuzzywuzzy for comparisons.