hejazn86 / NLP_IMDB_reviews

an Natural Language Process model using IMBD_reviews_dataset

NLP_IMDB_reviews

Building a Natural Language Process model by using IMDB_reviews_dataset

The purpose of the code:

• To practice the skills I have earned in Natural Lnguage Process as a branch of time_series field
• Build a NLP model step by step

Preparing the data

Loading the dataset

In this code we use a dataset from Tensorflow_datasets called IMDB_reviews, which contains 25000 movie reviews. To use this dataset, we first import the libraries we need: tensorflow and tensorflow_datasets

Then we load the datasets from tensorflow_datsets by using the code bellow:

imdb, info = tfds.load("imdb_reviews", with_info= True, as_supervised=True)

And split the data to a train_set and a test_set. Next we use a for loop to iterate over the train_set and extract the sentences of movie reviews and their labels from it and add them in list seperately, a list for the sentences and a list for the labels as it is shown in the following code:

train_sentences = []
train_labels = []

for s, l in train_set :
  train_sentences.append(s.numpy().decode('utf8'))
  train_labels.append(l.numpy())

We do the same thing for the test_set as well, then conver these list to numpy arrays for convinience.

train_sentences = np.array(train_sentences)
test_sentences = np.array(test_sentences)
train_labels = np.array(train_labels)
test_labels = np.array(test_labels)