Hacker News new | comments | show | ask | jobs | submit login

it's not that hard. Those that work with classifiers, this kind of thing is pretty easy. Identifying sarcasm and irony are hard, but 'nasty comments' can be identified pretty simply using the well known text classifier algorithms. You find the training data and use it to train something like an SVM.

I'd be surprised if pg hasn't experimented with it at least a bit given his history with text classification algorithms.

Guidelines | FAQ | Support | API | Security | Lists | Bookmarklet | Legal | Apply to YC | Contact