NLP Sentiment Analysis 3 — Questions and Answers
Question 1: What is 'subjectivity detection' as a preprocessing step in sentiment analysis?
- Identifying the author of a text
- Distinguishing objective factual statements from subjective opinionated text (Correct answer)
- Measuring the reading level of text
- Detecting the language of a document
Correct answer: Distinguishing objective factual statements from subjective opinionated text
Subjectivity detection filters out objective sentences so sentiment analysis focuses only on opinion-bearing content.
Question 2: Which technique helps capture long-range dependencies in sentiment classification using deep learning?
- Bag-of-words model
- Naive Bayes classifier
- LSTM or Transformer-based model (Correct answer)
- TF-IDF vectorization
Correct answer: LSTM or Transformer-based model
LSTMs and Transformers maintain context across long sequences, capturing dependencies that bag-of-words models miss.
Question 3: What does 'opinion mining' typically extract beyond overall sentiment polarity?
- The author's identity and publication date
- Opinion holders, opinion targets, and sentiment expressions (Correct answer)
- Grammar errors and spelling mistakes
- Document similarity scores
Correct answer: Opinion holders, opinion targets, and sentiment expressions
Opinion mining identifies who holds the opinion, what the opinion is about, and the expressed sentiment toward it.
Question 4: How does 'contextualized word embedding' (e.g., BERT) improve sentiment analysis over static embeddings?
- It reduces the vocabulary size needed for training
- It generates different vector representations for the same word depending on its context (Correct answer)
- It eliminates the need for labeled training data
- It automatically detects the sentiment language
Correct answer: It generates different vector representations for the same word depending on its context
Contextualized embeddings like BERT encode surrounding context, so 'great' in a positive vs. sarcastic sentence gets different representations.
Question 5: Which strategy is commonly used to handle sarcasm detection in sentiment analysis?
- Removing all exclamation marks from text
- Incorporating incongruity signals between sentiment and context or using multimodal cues (Correct answer)
- Applying a standard negation handler
- Using a simple polarity lexicon
Correct answer: Incorporating incongruity signals between sentiment and context or using multimodal cues
Sarcasm often involves contrast between literal positive words and negative context, so detecting incongruity is a key strategy.
Question 6: What is the primary limitation of lexicon-based sentiment analysis?
- It requires large amounts of labeled training data
- It struggles with domain-specific language and context-dependent word meanings (Correct answer)
- It cannot handle binary (positive/negative) classification
- It is computationally too expensive for real-time use
Correct answer: It struggles with domain-specific language and context-dependent word meanings
Lexicon-based methods rely on fixed word-sentiment mappings that fail when domain context shifts the meaning (e.g., 'sick' meaning 'cool' in slang).
Question 7: In multi-class sentiment analysis, what does 'fine-grained' classification refer to?
- Analyzing only short tweets rather than long reviews
- Classifying sentiment into more than three categories such as very positive, positive, neutral, negative, very negative (Correct answer)
- Performing sentiment analysis at the word level
- Distinguishing between multiple languages
Correct answer: Classifying sentiment into more than three categories such as very positive, positive, neutral, negative, very negative
Fine-grained sentiment goes beyond positive/negative/neutral to capture a spectrum of intensity levels.
NLP Sentiment Analysis 3
What is 'subjectivity detection' as a preprocessing step in sentiment analysis?