In this chapter, we covered basic RNN cells, LSTM cells, and the seq2seq model in building a language model that can be used for multiple NLP tasks. We implemented a chatbot, from scratch, to answer questions by generating a sequence of words from the provided dataset.
The experience in this exercise demonstrates the value of LSTM as an often necessary component of the RNN. With the LSTM, we were able to see the following improvements over past CNN models:
- The LSTM was able to preserve state information
- The length of sentences for both inputs and outputs could be variable and different
- The LSTM was able to adequately handle complex context
Specifically, in this chapter, we did the following:
- Gained an intuition about the RNN and its primary forms
- Implemented a language model using RNN
- Learned about the LSTM model
- Implemented the LSTM language model and compared it...
