Sentiment classification is a many-to-one RNN task. With input text, the task is to predict what is the sentiment of the text.
Below is an example where we try to output rating from a restaurant review.
This is our RNN architecture;
We can use word embeddings to boost up the algorithm’s performance. Also, even when an unseen word appears in the test time, if that word has a pre-trained word embedding, our algorithm can output descent prediction.