sharpening usually refers to "USM" (or "Unsharp Mask") technique, which modifies the pixels either side of a change in brightness, making the dark side of the change slightly darker and the bright side slightly brighter. This has the optical effect of sharpening the edges as in their unsharpenned form the jpeg pixels tend to blur the edge over a short distance of 2 or 3 pixels (or more) as they are square and rarely line up with the edge.
The amount of brightening/darkening, the distance from the edge over which it is applied, how much of a change in brightness is considered as an edge, and how often it is done all influence the amount of sharpenning perceived.
Amount of sharpenning depends on final image resolution and display size, also any resizing of resolution will destroy previous sharpenning. It should be done last in any processing sequence.
If over applied the sharpenning results in a halo effect around high contrast edges, it can also accentuate any noise in high ISO shots. My rule for sharpenning is only just enough then a little bit less.
It is a destructive one way process, the original image is correct but is perceived as "soft", sharpenning changes the jpeg data, every sharpenning moves the data further from the original data. So if you are going to keep original files and then post process them the most untreated version would be best.
You'll have to check the manual to see which one to use, as the "soft" setting may apply a blurr effect to produce a soft focus result, or it may be the unsharpenned image.