> The researchers were able to perform text classification using only text compression and a clustering algorithm (kNN).

Wasnt this result disproved, because the positive results were due to a bad kNN implementation? I recall reading something like this but can't recover the exact post/article...