What chunking strategies exist beyond fixed-size?
hardAnswer
- (1) Fixed-size (300-1000 tokens with overlap) — baseline.
- (2) Recursive splitting: split by paragraph → sentence → words until under max size, preserving structure.
- (3) Semantic chunking: use an embedding model to detect semantic boundaries (Kamradt) — chunks split where similarity between adjacent sentences drops.
- (4) Document-structure chunking: use headings / markdown structure as boundaries.
- (5) Late chunking (Günther 2024): embed the full document, then chunk the embedded tokens post-hoc — preserves cross-chunk context.
Check yourself — multiple choice
- Only fixed-size chunks
- Fixed-size, recursive, semantic (embedding boundaries), structure-based, late chunking (embed then split)
- Only one chunk per doc
- Chunk size doesn't matter
Chunking: fixed / recursive / semantic / structure / late chunking — each with trade-offs.
#rag#retrieval
Practise LLMs & GenAI
214 interview questions in this topic.