Domain-specific sentiment in NLP - Practice Problems & Coding Challenges

Challenge - 5 Problems

🧠 Conceptual

intermediate

Why is domain-specific sentiment analysis important?

Imagine you want to analyze customer reviews about medical devices. Why might a general sentiment model fail here?

A Because general sentiment models always perform better on any domain without retraining.

B Because domain-specific sentiment ignores the context of words and focuses only on word frequency.

C Because words like 'positive' or 'negative' have different meanings in medical contexts compared to general language.

D Because medical reviews are always neutral and don't need sentiment analysis.

💡 Hint

Think about how words can change meaning depending on the topic.

In domain-specific sentiment, words may have different sentiment meanings. For example, 'positive' in medical tests can mean a bad result, unlike general positive feelings.

❓ Predict Output

intermediate

Output of domain-specific sentiment prediction code

What is the output of this Python code that predicts sentiment using a domain-specific dictionary?

NLP

domain_sentiment_dict = {'stable': 1, 'critical': -1, 'improved': 1, 'declined': -1}
text = 'The patient condition is stable but later declined'
sentiment_score = sum(domain_sentiment_dict.get(word, 0) for word in text.lower().split())
print(sentiment_score)

A 2

B 0

C -1

D 1

💡 Hint

Check which words in the text match the dictionary and sum their scores.

'stable' has score 1, 'declined' has score -1, others 0, so total is 1 + (-1) = 0.

❓ Model Choice

advanced

Choosing a model for domain-specific sentiment analysis

You want to build a sentiment model for financial news articles. Which model choice is best?

A Use a rule-based sentiment model designed for movie reviews.

B Train a sentiment model from scratch using only financial news labeled data.

C Use a pre-trained general sentiment model without any fine-tuning.

D Fine-tune a pre-trained language model on labeled financial news sentiment data.

💡 Hint

Consider leveraging existing knowledge and adapting it to your domain.

Fine-tuning a pre-trained model on domain data combines general language understanding with domain-specific sentiment knowledge, improving accuracy.

❓ Metrics

advanced

Evaluating domain-specific sentiment model performance

You trained a domain-specific sentiment classifier. Which metric best shows how well it distinguishes positive and negative sentiment?

A F1-score

B Accuracy

C Mean Squared Error

D Perplexity

💡 Hint

Think about metrics that balance precision and recall for classification.

F1-score balances precision and recall, making it suitable for sentiment classification especially with imbalanced classes.

🔧 Debug

expert

Debugging domain-specific sentiment model training code

What error does this code raise when training a sentiment model with domain-specific data?

NLP

from sklearn.feature_extraction.text import CountVectorizer
from sklearn.linear_model import LogisticRegression
texts = ['good profit', 'bad loss', 'stable growth']
labels = [1, 0]
vectorizer = CountVectorizer()
X = vectorizer.fit_transform(texts)
model = LogisticRegression()
model.fit(X, labels)

A ValueError: Found input variables with inconsistent numbers of samples

B TypeError: 'int' object is not iterable

C AttributeError: 'CountVectorizer' object has no attribute 'fit_transform'

D No error, model trains successfully

💡 Hint

Check if the number of texts matches the number of labels.

There are 3 texts but only 2 labels, causing a mismatch error during training.