Recovering from Privacy-Preserving Masking with Large Language Models | IEEE Conference Publication | IEEE Xplore