I was impressed with the simplicity of some algorithms for natural language processing, such as a spelling for example.

These algorithms could help create processes/models to localize repetitive patterns in TXT files, and increasing the compression.


http://norvig.com/spell-correct.html