How does Sprout determine sentiment?
Sprout’s sentiment analysis is built using a machine learning technique called a Deep Neural Network (DNN). Deep Neural Networks accurately determine which parts of text are more or less important in determining a classification like sentiment. For example, in the sentence "I really loved the pizza," the word "loved" is the most important word that determines sentiment.
When a DNN collects enough data and tags as positive, negative or neutral, the DNN can automatically figure out which words or phrases are the most relevant and ignore the rest. A DNN can learn a language’s nuance and structure with very little additional information from humans.
Positive, Negative or Neutral?
When Sprout receives a block of text, the DNN computes a probability score for the positive, negative and neutral labels. Sprout then selects the label with the highest probability.
For example, for the text "I thought the movie was going to be terrible but I was pleasantly surprised by how good it was," the DNN computes 3 probabilities:
In this example, you can think of the model saying, "There’s a tiny chance this is neutral and a small chance that this text is negative, but I’m fairly certain it’s positive." Sprout selects the highest probability and returns that this text is positive.
Note: The model returns "unclassified" for languages that aren't supported and for inputs that don't contain any text.