A many-to-one architecture, as the name suggests, takes a sequence of input and maps it to a single output value. One such popular example of a many-to-one architecture is sentiment classification. A sentence is a sequence of words, so on each time step, we pass each word as input and predict the output at the final time step.

Let's suppose that we have a sentence: Paris is a beautiful city. As shown in the following figure, at each time step, a single word is passed as an input, along with the previous hidden state; and, at the final time step, it predicts the sentiment of the sentence: