Module 04 Knowledge Check
5 questions. Pass mark 4 out of 5. Answer every question before checking the answer key below, then retry after reading the feedback.
1. Splitting a document at a fixed character count risks…
- A. Larger indexes
- B. Cutting mid-sentence or mid-table so the chunk loses its meaning
- C. Slower embedding
- D. Nothing
2. PDF extraction commonly fails on…
- A. Plain paragraphs
- B. Multi-column layouts, tables, and scanned pages
- C. Headings
- D. Short files
3. Chunk metadata should record…
- A. Only the text
- B. Source, section, and identifiers needed for citation, filtering, and permissions
- C. The embedding
- D. The model
4. Changing your embedding model requires…
- A. No action
- B. Re-embedding the entire corpus, because vectors are not comparable across models
- C. Only new documents
- D. A prompt change
5. Chunk overlap exists mainly to…
- A. Increase index size
- B. Avoid losing an answer that straddles a split boundary
- C. Improve embeddings
- D. Reduce cost
---
Answer Key and Explanations
Check these only after attempting every question.
1. B - Cutting mid-sentence or mid-table so the chunk loses its meaning
Structure-aware splitting (headings, paragraphs) preserves meaning far better than blind character counts.
2. B - Multi-column layouts, tables, and scanned pages
Always inspect extracted text before indexing; garbled tables silently poison retrieval.
3. B - Source, section, and identifiers needed for citation, filtering, and permissions
Without provenance you cannot cite, filter, or later delete the right content.
4. B - Re-embedding the entire corpus, because vectors are not comparable across models
Mixed-model vectors produce silently meaningless similarity scores.
5. B - Avoid losing an answer that straddles a split boundary
It trades storage for a lower chance of the key sentence being cut in half.
Practical Check
Apply this module to your own work: complete the module activity for *Document Loading, Splitting, and Embeddings*, then write one sentence naming what your result shows and one naming what it does not.
Strong Answer Pattern
A strong answer names the task, the evidence used, the check performed, and the remaining limitation. It avoids "proved", "guaranteed", or "always" unless the evidence genuinely supports it.