Need ideas for selective batch process

[NEXUS77]

Can anyone help me figure this one out.
I have hundreds of dialogue regions in a project that I want to increase gain for selectively. They are dialogue and most were recorded at decent levels, but about thirty percent are just too low. I want to bump their gain up a few db but I don't want to adjust the gain of the louder files because I want to generally keep their relative levels as they are. (so that soft speaking isn't the same level as screaming. I basically would like to be able to say, "If a file falls within this db range, add 4 db of gain." I would also like to do the same for the higher ranges and just normalize them. Any thoughts or ideas would be greatly appreciated. The only thin I can think of is creating limiting and compression, but I don't really want to squash dynamic and unnecessarily raise noise floors. Thanks!!