Sentiment with context (sarcasm, negation) in NLP - Practice Problems & Coding Challenges

Challenge - 5 Problems

❓ Predict Output

intermediate

Output of sentiment prediction with negation handling

Given the following Python code snippet using a simple rule-based sentiment function that handles negation, what is the output when the input sentence is "I do not like this movie"?

NLP

def simple_sentiment(text):
    positive_words = {'like', 'love', 'good', 'great', 'happy'}
    negative_words = {'hate', 'bad', 'terrible', 'sad', 'dislike'}
    words = text.lower().split()
    negation_words = {'not', 'no', 'never'}
    negation = False
    score = 0
    for word in words:
        if word in negation_words:
            negation = True
            continue
        if word in positive_words:
            score += -1 if negation else 1
            negation = False
        elif word in negative_words:
            score += 1 if negation else -1
            negation = False
    if score > 0:
        return 'Positive'
    elif score < 0:
        return 'Negative'
    else:
        return 'Neutral'

print(simple_sentiment("I do not like this movie"))

A Positive

B Negative

C Neutral

D SyntaxError

💡 Hint

Think about how negation flips the sentiment of the word 'like'.

The word 'like' is positive, but it is preceded by 'not', a negation word. The code flips the sentiment score for 'like' to negative, resulting in an overall negative sentiment.

🧠 Conceptual

intermediate

Understanding sarcasm detection challenges

Why is sarcasm particularly difficult for sentiment analysis models to detect?

A Because sarcasm is only present in spoken language, not in text.

B Because sarcasm always uses complex vocabulary that models cannot understand.

C Because sarcasm often uses positive words to express negative feelings, confusing simple word-based models.

D Because sarcasm depends on the length of the sentence, which models ignore.

💡 Hint

Think about how the literal words differ from the intended meaning.

Sarcasm often uses positive words ironically to express negative feelings, which simple models that rely on word sentiment struggle to interpret correctly.

❓ Metrics

advanced

Evaluating sarcasm detection model performance

A sarcasm detection model was tested on 1000 sentences. It correctly identified 80 sarcastic sentences out of 100 sarcastic ones, and correctly identified 890 non-sarcastic sentences out of 900 non-sarcastic ones. What is the model's precision for the sarcastic class?

A 0.89

B 0.50

C 0.44

D 0.80

💡 Hint

Precision = True Positives / (True Positives + False Positives). Calculate false positives first.

True Positives (TP) = 80, False Negatives (FN) = 20 (100-80), True Negatives (TN) = 890, False Positives (FP) = 10 (900-890). Precision = 80 / (80 + 10) = 80 / 90 ≈ 0.89.

🔧 Debug

advanced

Debugging a sarcasm detection model code snippet

What error will the following Python code raise when run? def detect_sarcasm(text): sarcasm_keywords = ['yeah right', 'as if', 'totally'] for phrase in sarcasm_keywords: if phrase in text.lower(): return True return False print(detect_sarcasm(12345))

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

B NameError: name 'text' is not defined

C TypeError: argument of type 'int' is not iterable

D No error, output is False

💡 Hint

Check the method called on the input and the input type.

The code calls text.lower(), but the input is an integer (12345), which does not have a lower() method, causing an AttributeError.

❓ Model Choice

expert

Choosing the best model architecture for sarcasm detection with context

Which model architecture is best suited to detect sarcasm in text by understanding context and subtle language cues?

A A k-nearest neighbors model using TF-IDF vectors

B A simple bag-of-words logistic regression model

C A convolutional neural network (CNN) with fixed word embeddings

D A transformer-based model like BERT fine-tuned on sarcasm-labeled data

💡 Hint

Consider models that capture word order and context deeply.

Transformer models like BERT understand context and subtle language patterns, making them best for sarcasm detection compared to simpler models.