Finding the right representation for your NLP data - Tryolabs Blog

When considering what information is important for a certain decision procedure (say, a classification task), there's an interesting gap between what's theoretically --that is, actually-- important on the one hand and what gives good results in practice as input to machine learning (ML) algorithms, on the other. Let's look at sentiment analysis tools as an example. Expression of sentiment is a pragmatic phenomenon. To predict it correctly, we need to know both the meaning of the sentences and the context in which those sentences appeared. How do you get the meaning of a sentence?