Abstractive vs Extractive Summarization Quiz

Answers and explanations appear as you take the quiz.

1. What does an extractive summarizer do?
   • Selects and copies the most important existing sentences
   • Rewrites the document in entirely new words
   • Translates the document into another language
   • Generates an image of the content
2. Which summarization type carries the higher risk of hallucinating facts?
   • Extractive
   • Abstractive
   • Both equally
   • Neither can hallucinate
3. Why can extractive summaries feel choppy or redundant?
   • They use too many emojis
   • They are translated badly
   • They stitch together independently selected sentences without rewriting transitions
   • They are always too short
4. Which models are commonly used for abstractive summarization?
   • TextRank and TF-IDF only
   • K-means clustering
   • Linear regression
   • BART, PEGASUS, and T5
5. What is PEGASUS's distinctive pretraining objective?
   • Masking and regenerating whole important sentences (gap-sentence generation)
   • Predicting the next pixel in an image
   • Sorting sentences alphabetically
   • Counting words in a document
6. Why might a legal team choose extractive summarization for a clause?
   • It guarantees that the selected clause is legally complete
   • It preserves selected source wording and avoids introducing paraphrase changes
   • It proves that the source document is accurate
   • It removes the need to review the surrounding context
7. What is one risk of extractive summaries even when they copy source sentences exactly?
   • They always generate words that do not appear in the source
   • They cannot contain any sentence from the document
   • Selection can remove context and make the result misleading
   • They necessarily translate every sentence
8. What does an extract-then-abstract summarization workflow do?
   • Copies every sentence with no selection
   • Generates text without consulting the source
   • Replaces the document with a keyword list
   • Selects source evidence first, then generates a concise summary from that evidence