Python Social Media Analytics

Labeling the data

As explained in the previous section, we have to manually label a training dataset. We assume that the sentiment is specific to the topic of analysis and that a person who labels the data is able to do it correctly. In order to store the labels, we create a column label, which associates a class with a tweet.

The general rule is that the more labeled data the better, but labeling is a costly operation. In order to be efficient we label only observations that are clearly associated with a class. It will help us to get only the best examples of positive, neutral, and negative sentiment. As a result, the algorithm should have a good performance in predicting the most insightful tweets in terms of sentiment, which is the objective. ...