no .. it can not be seperated by removing the sound and leaving the words. Its because of the way the singing/words are incorporated into the music . Many moons ago , they used to use left bank verses right bank for serperating music from vocals , ( like some old Sonny and Cher music and the original Chipmunks Christmas song ) , but not any more .
Can I ask , what are you needing the word part seperated for ?
If its for an animated talking skull , or something similiar , simply use a program like AUDACITY and mix the 2 channels together , then blank out one of the channels ( say you take back channels and mix them onto the RIGHT channel , then take the LEFT channel and blank it ouot so nothings on it )
Then use your mic and basically , sing to the song but make sure your voice is going onto a seperate NEW channel first and then copy and paste your singing , onto the LEFT channel and feed STRICKLY the left channel into the skull and remove the speaker from it , so your voice is not heard and then allow the music from the RIGHT channel to be heard by the audience .
This way , the skulls mouth will move ONLY to the signing ( via your voice on the RIGHT channel ) and the audience will ONLY hear the original music/singing on the LEFT channel . And the skulls mouth movement will be in perfect time with JUST the singing parts of the song . And FTR - it doesn't matter how lousey you sing , no one is gonna hear it anyway - just the cicuit board for the talking skull - LOL