Sentiment Analysis using VADER - Using Python

Sentiment analysis is one of the most common Natural Language Processing (NLP) tasks used to determine the emotional tone of a piece of text. It helps identify whether a sentence expresses a positive, negative, or neutral opinion, making it useful for analyzing customer reviews, social media posts, product feedback, and online discussions.

• Among the various sentiment analysis techniques, VADER (Valence Aware Dictionary and sEntiment Reasoner) is a popular rule-based approach that is specifically designed for short and informal text.
• It uses a predefined sentiment lexicon along with linguistic rules to understand the sentiment of a sentence, while also considering factors such as emojis, punctuation, capitalization, negation words, and intensifiers.

Working

1. Assigns sentiment scores: Matches words against a predefined sentiment lexicon to determine their emotional polarity.
2. Applies linguistic rules: Adjusts scores based on negation words, punctuation, capitalization, conjunctions, and intensifiers.
3. Calculates sentiment metrics: Computes positive (pos), negative (neg), neutral (neu), and compound scores for the input text.
4. Classifies the sentiment: Uses the compound score to determine whether the overall sentiment is positive, negative, or neutral.
5. Uses predefined thresholds: A compound score ≥ 0.05 indicates positive sentiment, ≤ -0.05 indicates negative sentiment, and values in between are considered neutral.

Sentiment Scores in VADER

VADER evaluates the sentiment of a sentence by returning four numerical scores that represent different aspects of the text. These scores help determine the overall sentiment expressed in the input.

Implementation of Sentiment Analysis using VADER

In this implementation, we will use the VADER SentimentIntensityAnalyzer to calculate the sentiment scores of different sentences.

Step 1: Install the Required Library

• Install the vaderSentiment library using the following command:
• vaderSentiment: A Python library that provides the VADER sentiment analysis model for analyzing the emotional tone of text.

!pip install vaderSentiment

Step 2: Import the Required Library

• Import the SentimentIntensityAnalyzer class.
• SentimentIntensityAnalyzer is the core class used to calculate sentiment scores for a given sentence.

from vaderSentiment.vaderSentiment import SentimentIntensityAnalyzer

Step 3: Create the Sentiment Analyzer

• Create an analyzer object that will be used throughout the program.
• Initializes the pretrained VADER sentiment analyzer.
• The same object can be reused to analyze multiple sentences.

sid_obj = SentimentIntensityAnalyzer()

Step 4: Create a Function to Analyze Sentiment

• Define a function that calculates the sentiment scores and determines the overall sentiment.
• polarity_scores() calculates the sentiment scores for the input text.
• Displays the positive, negative, neutral, and compound scores.
• Uses the compound score to classify the overall sentiment as Positive, Negative, or Neutral.

def sentiment_scores(sentence):

    sentiment_dict = sid_obj.polarity_scores(sentence)

    print("Sentiment Scores:", sentiment_dict)
    print(f"Negative Sentiment: {sentiment_dict['neg']*100:.1f}%")
    print(f"Neutral Sentiment: {sentiment_dict['neu']*100:.1f}%")
    print(f"Positive Sentiment: {sentiment_dict['pos']*100:.1f}%")

    if sentiment_dict["compound"] >= 0.05:
        print("Overall Sentiment: Positive")
    elif sentiment_dict["compound"] <= -0.05:
        print("Overall Sentiment: Negative")
    else:
        print("Overall Sentiment: Neutral")

Step 5: Analyze Sample Sentences

• Call the function with different input sentences.
• The first two sentences are expected to produce a positive sentiment.
• The third sentence contains negative words, resulting in a negative sentiment.

if __name__ == "__main__":

    print("\n1st Statement:")
    sentiment_scores(
        "GeeksforGeeks is an excellent platform for learning programming."
    )

    print("\n2nd Statement:")
    sentiment_scores(
        "The presentation was informative and well organized."
    )

    print("\n3rd Statement:")
    sentiment_scores(
        "I am feeling disappointed with today's results."
    )

Output:

You can download the code from here.

Applications

• Social Media Monitoring: Analyzes posts, tweets, and comments to understand public opinion about brands, events, or products.
• Customer Feedback Analysis: Identifies positive and negative sentiments in product reviews and customer feedback to improve services.
• Brand Reputation Management: Tracks online sentiment to detect changes in customer perception and respond to negative feedback quickly.
• Market Research: Measures consumer opinions on products, advertisements, and marketing campaigns using sentiment trends.
• Review Classification: Automatically categorizes movie, hotel, restaurant, or e-commerce reviews based on sentiment.
• Chat and Comment Analysis: Monitors user comments, discussion forums, and chat messages to identify overall user satisfaction.

Advantages

• Uses a predefined sentiment lexicon, eliminating the need for labeled training data.
• Effectively analyzes tweets, reviews, and short informal text containing slang and abbreviations.
• Considers emojis, capitalization, punctuation, negation, and degree modifiers while computing sentiment.
• Provides sentiment scores with low computational overhead, making it suitable for real-time applications.
• Can be incorporated into Python applications with minimal code and dependencies.
• Returns separate positive, negative, neutral, and compound scores, making sentiment predictions easy to understand.

Limitations

• May struggle with sentences that require deep contextual or domain-specific knowledge.
• Words not present in the sentiment dictionary may not contribute accurately to the final sentiment score.
• Primarily designed for short text and may not perform consistently on lengthy articles or reports.
• Optimized mainly for English text and requires additional resources for multilingual sentiment analysis.
• Being rule-based, it cannot improve automatically from new data unlike machine learning or transformer-based models.

Related Articles

• Sentiment Analysis using TextBlob in Python
• Text Preprocessing Techniques in NLP