script.py

'''
Firt, import the required packages
'''
import os
import emoji
import keras_nlp
import config

os.environ["KERAS_BACKEND"] = config.keras_backend

'''
This Basic Sentiment Analysis is based on a pretrained backbone taken from keras_npl models
'''
classifier = keras_nlp.models.BertClassifier.from_preset("bert_tiny_en_uncased_sst2")

'''
Now the user should type in the sentence that it will be analysed
'''
sentence = input("Type a sentence: ")
sentiment_matrix = classifier.predict([sentence])

'''
This is to print the result in terms of a 1x2 polarization matrix
'''
print("The result of this binary classification is {} negative and {} positive".format(sentiment_matrix[0,0], sentiment_matrix[0,1]))

'''
The following lines are just to transform numbers into emojis. The limits are only based on experience.
'''
if (sentiment_matrix[0,0] >= 1.8):
    print(emoji.emojize('Therefore, this sentence is classified as very negative :thumbs_down: :thumbs_down:'))
elif (sentiment_matrix[0,0] >= 0.5):
    print(emoji.emojize('Therefore, this sentence is classified as negative :thumbs_down:'))
elif (sentiment_matrix[0,0] >= -0.5):
    print(emoji.emojize('Therefore, this sentence is classified as neutral :man_shrugging:'))
elif (sentiment_matrix[0,0] >= -1.8):
    print(emoji.emojize('Therefore, this sentence is classified as positive :thumbs_up:'))
else: #sentiment_matrix[0,0] < -1.8
    print(emoji.emojize('Therefore, this sentence is classified as very positive :thumbs_up: :thumbs_up:'))