Most dialogue is in the center channel, unless the scene requires otherwise, and effects tend to go to the side/rears unless, again, the scene requires otherwise. You want to use something like Besweet (free) to split up the AC3 file.

If a person's voice pans from center to, say, left front (due to walking or camera movement), then you'll have to take both tracks and combine them. If there are effects in the center channel with dialogue, you probably won't be able to to remove them entirely without affecting the voice (a gunshot, for example.) Normal human speech is generally between 20-20,000 Hz, so you need to set up your audio program to drop everything above and below that (think equalizer), and see what kind of results that gets you.