Good point about introducing different spelling and word substitution. The original approach is for general data processing. If text document is the only kind

Good point about introducing different spelling and word substitution. The original approach is for general data processing. If text document is the only kind of data, then add a pre-processing step to normalize the text. Remove all the stop words. Normalize words to their root to smooth out the different spellings and substitution. Even the Different word normalizations can be tried, Soundex, NYSIIS, or Metaphone.