Skip to content
Related Articles

Related Articles

Sentiments in Text – Word Based Encodings
  • Difficulty Level : Medium
  • Last Updated : 10 Jul, 2020

Sentimental analysis is the processing of describing whether a particular feeling or an opinion is positive, negative, or neutral. For example, ” I hate my Lunch”, “I love my Lunch” and “I am fine with my Lunch”.There is a negative, positive, and neutral tone in each of these sentences. On a large scale, Sentimental Analysis is used in determining the customer’s feedback through comments. These comments help in building recommendation systems for future reference.

How to get the meaning of a word in a sentence?

We could have taken ASCII values of a character, but would that help us understand the semantics of a word?. Let us take the word “binary” into consideration, it can also be written as “brainy”. Clearly, both of these words share the same ASCII value but has a different meaning altogether. It is a tough task to train a neural network with words. The solution to all of this is if we could give words value and use them in the training model

Consider this sentence ” I love my lunch” let us give some random number to it. Let’s say the values are 1, 2, 3, and 4 respectively. Let’s say we have another sentence which is “I love my cat”, we can reuse the previous values and give a new token to the word “cat”. Let’s say the value of the cat is 5. There is a similarity of 4 values in both of those sentences. It’s a start on how to train a neural network. Fortunately, we have API’s like Tensorflow. Follow the below steps to train your model

  • Step1: Importing required Libraries

  • Step2: Create a List of sentences

  • Step3: Create a Tokenizer object

  • Step4: Use fit_on_text method

  • Step5: print out the word_index

Below is the implementation.

import tensorflow as tf
from tensorflow import keras
from tensorflow.keras.preprocessing.text import Tokenizer
sentences = [
    "I love my lunch",
    "I love my cat !"
tokenizer = Tokenizer(num_words = 100)
word_index = tokenizer.word_index


{'i': 1, 'love': 2, 'my': 3, 'lunch': 4, 'cat': 5}


My Personal Notes arrow_drop_up
Recommended Articles
Page :