The Holy Grail of text classification research is not to train a 100% accurate classification system, rather, it is to understand our own thought process better to produce consistent manually labeled data to train the system on.
In other news,
my thesis is back from the two external examiners with good positive reviews and recommendations of accept. Yay!
Refer the post on the long drawn out process.