Another thought is to analyse the data using weighted upvotes and downvotes.
For example for each vote vote cast, tally:
5/5 => 2 votes, 4/5 => 1 vote, 3/5 => 0 votes (neutral), 2/5 => -1 votes,1/5 => -2 votes.
Then for each item:
popular items get a high positive score
unpopular items get a negative score
contentious items have high variance
the number of votes indicates the level of interest