November 2019
Intermediate to advanced
346 pages
9h 36m
English
Start by collecting a dataset of reviews you'd like to imitate (step 1). A practical example would require a large corpus of reviews. There are many such datasets available, such as the Yelp reviews dataset. Proceeding to step 2, we create a mapping between characters and numbers. This will allow us to vectorize the text. Depending on your application, you may want to use the standard ASCII code. However, if you are using only a small number of characters, then this will unnecessarily slow down your model. We go on to declare the architecture of an RNN to learn and predict the sequence of characters (step 3). We used a relatively simple architecture. As will be shown in the next section, it nonetheless provides convincing results. ...