Lexicon-based approaches (VADER) in NLP - Practice Problems & Coding Challenges

Challenge - 5 Problems

❓ Predict Output

intermediate

VADER Sentiment Polarity Scores Output

What is the output of the following Python code using VADER sentiment analyzer?

NLP

from vaderSentiment.vaderSentiment import SentimentIntensityAnalyzer
analyzer = SentimentIntensityAnalyzer()
sentence = "I love sunny days but hate the rain."
scores = analyzer.polarity_scores(sentence)
print(scores)

A {"neg": 0.292, "neu": 0.458, "pos": 0.25, "compound": -0.0516}

B {"neg": 0.25, "neu": 0.5, "pos": 0.25, "compound": 0.0}

C {"neg": 0.0, "neu": 0.5, "pos": 0.5, "compound": 0.6249}

D {"neg": 0.292, "neu": 0.458, "pos": 0.25, "compound": 0.6249}

💡 Hint

Remember that VADER returns four scores: negative, neutral, positive, and compound.

VADER analyzes the sentence and assigns scores for negative, neutral, positive, and a compound score summarizing overall sentiment. The sentence has both positive and negative parts, so the compound score is close to zero but slightly negative.

🧠 Conceptual

intermediate

Understanding VADER's Compound Score Range

What is the range of the compound score produced by VADER sentiment analysis, and what does it represent?

A Range is from 0 to 100; it represents the percentage of positive words in the text.

B Range is from 0 to 1; it represents the probability of positive sentiment.

C Range is from -1 to 0; it represents the intensity of negative sentiment only.

D Range is from -1 to 1; it represents the overall sentiment from most negative to most positive.

💡 Hint

Think about how VADER summarizes sentiment in one number.

The compound score is a normalized value between -1 (most negative) and 1 (most positive) summarizing the overall sentiment of the text.

❓ Metrics

advanced

Evaluating VADER Sentiment Classification Accuracy

You have a dataset of 1000 sentences labeled as positive or negative. Using VADER's compound score with a threshold of 0.05 for positive and -0.05 for negative, which metric best measures how well VADER classifies sentiment?

A Accuracy, because it measures the proportion of correctly classified sentences.

B Silhouette Score, because it measures cluster separation.

C Mean Squared Error, because it measures the difference between predicted and true labels.

D Perplexity, because it measures language model uncertainty.

💡 Hint

Consider the task is classification, not regression or clustering.

Accuracy is the appropriate metric for classification tasks to measure the proportion of correct predictions.

🔧 Debug

advanced

Identifying Error in VADER Usage Code

What error will the following code raise when run? from vaderSentiment.vaderSentiment import SentimentIntensityAnalyzer analyzer = SentimentIntensityAnalyzer() scores = analyzer.polarity_scores(12345) print(scores)

A ValueError: invalid literal for int() with base 10

B TypeError: expected string or bytes-like object

C No error; outputs sentiment scores

D AttributeError: 'int' object has no attribute 'lower'

💡 Hint

VADER preprocesses input by converting it to a string first.

No error is raised. VADER's polarity_scores converts non-string inputs to strings using str() before processing. '12345' contains no sentiment words, yielding neutral scores.

❓ Model Choice

expert

Choosing the Best Sentiment Analysis Approach for Social Media Text

You want to analyze sentiment of short social media posts with slang, emojis, and informal language. Which approach is best suited?

A Traditional bag-of-words model with logistic regression, because it captures word frequency.

B Lexicon-based approach using VADER, because it is tuned for social media text and handles emojis and slang well.

C Topic modeling with LDA, because it finds themes in text.

D Rule-based sentiment analysis using fixed dictionaries without handling slang.

💡 Hint

Consider which method is designed for informal, short texts with emojis.

VADER is specifically designed for social media text and handles slang, emojis, and informal language better than traditional models.