Updated the parse_output function in metadata_preproc.py to ensure that empty strings and single-space characters are not captured as valid words during post-processing. This prevents downstream issues where noise or whitespace might be incorrectly interpreted as content tokens.