1Z0-184-25 Question 106
Single answerYou are building a product recommendation system for an e-commerce client who wants to store and query vector embeddings for product descriptions in Oracle Database 23c. The client aims to avoid transferring data outside the database and reduce latency by generating these embeddings directly in the database. Which approach best meets these requirements?
- A
Leverage Oracle Machine Learning for Python with a pre-trained text embedding model, generate embeddings in-database, store them in a VECTOR type column, and create a vector index for querying.
- B
Export product descriptions to a CSV file, generate embeddings in an external environment, then import the embeddings daily into a BLOB column.
- C
Use Oracle Advanced Compression to compress textual data and rely on standard B-tree indexes for semantic similarity searches.
- D
Configure Oracle Spatial geometry columns to store text embeddings and use spatial operators for similarity queries.
Show answer and explanation
Correct answer: A
Explanation
Oracle Database 23c introduces a vector data type and indexing strategy that, combined with Oracle Machine Learning for Python, enables efficient generation and querying of embeddings in-database. By generating embeddings directly in the database, you minimize data movement, reduce latency, and can perform vector-based similarity queries with fewer integration points. For more details, refer to Oracle Database 23c documentation on vector search and Oracle Machine Learning for Python.
- A. Correct.
Correct. Oracle Machine Learning for Python allows the generation of text embeddings directly in the database without exporting data. By storing embeddings in a VECTOR column and indexing them with a vector index, you can perform efficient similarity queries.
- B. Incorrect.
Incorrect. Although generating embeddings offline is feasible, it contradicts the goal of minimizing data movement and latency. Regularly exporting and importing data adds overhead and can delay real-time recommendations.
- C. Incorrect.
Incorrect. Advanced Compression reduces storage costs for textual data but does not provide a vector-based search mechanism for semantic queries. Compression alone does not transform text into embedding vectors.
- D. Incorrect.
Incorrect. Oracle Spatial is designed for geometry and geospatial data, not for text embeddings. Using spatial operators for semantic analysis of text is not a recommended approach.