Emotion classification using NRC Lexicon in Python

Emotion identification or recognition is the ability of a person or an object to sense a particular emotion exhibited in the environment and place it into one of the many categories of emotions.

Emotion Classification in python is a feasible alternative to the traditional sentiment analysis technique which labels words or sentences as positive or negative and allots them with a polarity score accordingly.

The basic ideology behind this algorithm is to mimic the human thought process and it tries to segment the words portraying emotions from the text. The analysis is performed with a training set of data where a set of pre assumed information is fed to the system which acts as the basis for classification.

Here's the package based on the WordNet synonym sets from the NLTK library and the affect lexicon of the National Research Council of Canada (NRC), which has over 27,000 terms.

The following categories are used by the library to measure and classify words' emotional effects −

• fear

• anger

• anticipation

• trust

• surprise

• positive

• negative

• sadness

• disgust

• joy

Steps for installation

• Step 1 − install the NRC module by using the pip install command in the terminal.

pip install NRCLex

Installing in a jupyter

notebook and command prompt usually follows the same steps if you’re using Windows.

Installing in MacOs also follows the same command. Use the terminal directly.

• Step 2 − install textblob in addition with nrclex to avoid facing MissingCorpusError

pip install textblob

• Step 3 −download corpora from textblob

python -m textblob.download_corpora

After installation, we can go ahead and import the library and create a text object.

Elementary Methods

1. raw text to filtered text(for best results, 'text' should be unicode).

text_object.load_raw_text(text: str)

2. converts tokenized list of words into a list of tokens

text_object.load_token_list(list_of_tokens: list)

3. Return words list.

text_object.words

4. Return sentences list.

text_object.sentences

5. Return affect list.

text_object.affect_list

6. Return affect dictionaries.

text_object.affect_dict

7. Return raw emotional counts.

text_object.raw_emotion_scores

8. Return highest emotions.

text_object.top_emotions

9. Return frequencies.

Text_object.frequencies

Here, we have made use of the top_emotions function to classify a list of words based on emotions.

Algorithm

Step 1 − import nrclex import nrclex

Step 2 − from nrclex import NRCLex

Step 3 − initialize a list of strings-words that you’d like to classify

Step 4 − for i in range len(text)

Step 4 − emotion = NRCLex(text[i]) #creates an object for each text

Step 5 − emotion.top_emotions #classifies the emotion

Example

# Import module
import nrclex
from nrclex import NRCLex

text = ['happy', 'beautiful', 'exciting', 'depressed']

# Iterate through list
for i in range(len(text)):

   # call by object creation
   emotion = NRCLex(text[i])

   # Classify emotion
   print('\n', text[i], ': ', emotion.top_emotions) 

Output

innocent : [('trust', 0.5), ('positive', 0.5)]
hate : [('fear', 0.2), ('anger', 0.2), ('negative', 0.2), ('sadness', 0.2), ('disgust', 0.2)]
irritating : [('anger', 0.3333333333333333), ('negative', 0.3333333333333333), 
('disgust', 0.3333333333333333)]
annoying : [('anger', 0.5), ('negative', 0.5)]

Algorithm

Step 1 − import nrclex

Step 2 − from nrclex import NRCLex

Step 3 − initialize a list of strings-words that you’d like to classify

Step 4 − for i in range len(text)

Step 4 − emotion = NRCLex(text[i]) #creates an object for each text

Step 5 − emotion.top_emotions #classifies the emotion

Example

import nrclex
from nrclex import NRCLex
 
# Assign list of strings
text = ['innocent','hate', 'irritating','annoying']
 
# Iterate through list
for i in range(len(text)):
 
   # Create object
   emotion = NRCLex(text[i])

   # Classify emotion
   print('\n\n', text[i], ': ', emotion.top_emotions) 

Output

innocent :  [('trust', 0.5), ('positive', 0.5)] 
 hate :  [('fear', 0.2), ('anger', 0.2), ('negative', 0.2), ('sadness', 0.2), ('disgust', 0.2)] 
irritating :  [('anger', 0.3333333333333333), ('negative', 0.3333333333333333), ('disgust', 0.3333333333333333)] 
 annoying :  [('anger', 0.5), ('negative', 0.5)] 

Conclusion

The NRC Emotion Lexicon is widely used in research and industry for sentiment analysis and emotion classification tasks. This means that there is a large community of users and resources available for support and further development. NRCLex has also managed to break the language barrier by providing a stable output for more than 100 languages worldwide with the help of google translate. This finds various applications in the health care sector in understanding pandemic responses. Practical applications include psychology and behavioral sciences, fake news detection and enhancing human-computer interaction.