5 stars are a very popular metaphor for visualising, very often sparkline-like, the popular opinion about an item. The values themselves usually come from voting, reviews and the like. The are popular because they show in a bar-graph like fashion what is the rating, and you can visually compare items. The downside is that as bar-graphs are quite crude, because they are one dimentional. It is not possible to judge if there is a consesus, or there are controversial opinions about the item in question. No idea about the distribution of data.
Can we design a 5-star rating to display the distribution of opinion, while preserving the sparkline properties of the original.?
Yes. The basic idea is to vary the size of the stars to display the different vote numbers in each category. The simplest (intuitive?) idea is that the bigger the star, the more votes in that category. The horizontal dimention remains the same as used by amazon, netflix, google, and the rest of the net.
How to size the stars should vary depending on context, application and the data itself. If the data allows it, keeping the areas of the stars related to the number of votes is probably best, but there is a danger that there will be a very big difference between the star sizes, which can lead to losing the word-like quality of the original star chart. If they are important, then the simplest is to drop the size-area relation, and reduce it to just preserve the order, the bigger stars, mean more votes.
