1.5.2 · Lesson 18 of 28
Embeddings
An embedding converts content into a vector representation. Conceptually: Semantically similar content produces vectors that can be compared for similarity.
WHAT
An embedding converts content into a vector representation.
Conceptually:
"Refund policy for premium customers"
↓
Embedding Model
↓
[0.12, -0.44, ...]Semantically similar content produces vectors that can be compared for similarity.
AWS
- Amazon Titan embedding models through Amazon Bedrock
- Lambda for batch processing
- S3 for source data
Selection factors
- domain fit
- dimensions
- quality
- latency
- cost
- language support
- scale
EXAM TRAP
Embedding model selection matters.
Changing the embedding model may require re-embedding existing content.
---
Key points
- An embedding converts content into a vector representation. Conceptually: Semantically similar content produces vectors that can be compared for similarity.
Exam focus
Embedding model selection matters. Changing the embedding model may require re-embedding existing content.