I've looked into this before. It can't be done effectively. In the same way that cutting out someone from a photograph doesn't let you see what was behind them at the time, you can't separate music from dialogue and hear them both separately as they were. Some clever software can try and separate frequencies and give you a rough approximation of the individual sounds but you can NEVER unmix sound that's already mixed. The only hope is that one day there'll be a release with just dialogue on the centre channel. Normally doesn't happen though. The rare exception was The Clone Wars CG series. That would be a fan editor's dream if it was actually worth editing.