You should be able to find a configuration by tilting / rotating the sensor so that when it sits still on your desk it only gives output on one channel (the other two reading zero, or very nearly zero). The channel with non-zero output is now measuring the gravitational acceleration due to the earth which is 9.8 m/s/s (or 1g, depending on the units you want to use), so now you have your calibration constant.

It's likely that each channel will have the same calibration constant, but you can rotate it to zero out different channels to verify the calibration for each.

EDIT: on closer inspection of the data sheet, it appears there is a bias voltage at 0g, so looking for a 0V output as I suggested would not work. You'll have to use the orientation diagram in the datasheet to align an axis with vertical, then record the non-vertical axes output as the offsets for those axes. Then rotate the device so a new axis is vertical, get that reading, subtract the offset you just measured, and now you can calibrate. (disclaimer: I haven't actually worked with these devices so there may be a gap in my understanding, but I'm sure someone more knowledgeable will jump in if I'm wrong.)