In simple words: An RNN reads in sequence and carries a changing memory state.
RNN basics
A Recurrent Neural Network (RNN) processes a sequence one step at a time. It carries a hidden state from one step to the next, so earlier information can affect later predictions.
The limitation
In a long sequence, the influence of distant information can fade during learning. RNNs can work for text, speech, and time series, but long dependencies are hard for a basic RNN. Long Short-Term Memory (LSTM) networks were designed to manage memory more effectively.
See the flow
word 1→state→word 2→state→word 3
Simple example
To interpret “The package was late, so I complained,” the model needs the order and relationship between the words.
Remember: An RNN reads in sequence and carries a changing memory state.