This question gets asked often, we ought to have a sticky. To the best of my knowledge:
With 5.1 audio, dialogue in the centre channel, it can be done with BeSweet by decoding into separate WAVs and extracting just the front left and front right channels. You will be left with music and effects, but dialogue should be mostly gone.
With 2-channel Dolby surround, you could try a 5.1 upmix (see Doom9) then discard the unwanted channels as above.
With stereo audio, it may be possible to reduce dialogue with an audio editor, e.g. Sound Forge, by the same vocal elimination techniques used to create karaoke tracks (LINK). Or you could try software such as THIS. Either way, the results are not likely to be perfect.
With mono audio, I don't think there's a chance in hell.