In the last section, we briefly learned about word embedding without implementing it. In this section, we will download a dataset called IMDB, which contains reviews, and build a sentiment classifier which calculates whether a review's sentiment is positive, negative, or unknown. In the process of building, we will also train word embedding for the words present in the IMDB dataset. We will use a library called torchtext that makes a lot of processes such as downloading, text vectorization, and batching much easier. Training a sentiment classifier will involve the following steps: 

1. Downloading IMDB data and performing text tokenization
2. Building a vocabulary
3. Generating batches of vectors
4. Creating a network model with embeddings
5. Training the model