On: Ask HN: Algorithm to identify the same events based on semantic similarity

[ntoshev]

Statistical language processing is a good way to go, but he must have enough data for it to be useful. I'd recommend to start by gathering the data.

[abhikshah → ntoshev]

Amazon's Mechanical Turk seems like a good way to gather training data.. and it might even be cheap enough to use instead of a trained model (at least at first).