1Z0-184-25 Question 12
Single answerYou are designing a product recommendation feature in an e-commerce application that runs on Oracle Database 23c in Autonomous Database on Oracle Cloud Infrastructure. The application uses text and image embeddings stored as vectors to provide users with highly relevant product suggestions in near real-time. Which approach ensures efficient storage of these embeddings and the ability to perform semantic vector searches for top-k similar items?
- A
A. Store embeddings as plain text in CLOB columns, then use a standard text index to retrieve top-k similar items.
- B
B. Create a dedicated NUMBER-type column for embeddings, manually encode each dimension, and run regular numeric comparisons.
- C
C. Store embeddings in Oracle� native vector data type columns, build a vector index, and use semantic top-k queries to find similar items.
- D
D. Use an external NoSQL database for all embeddings and periodically batch-insert them into Oracle Database 23c for semantic queries.
Show answer and explanation
Correct answer: C
Explanation
Oracle Database 23c includes a native vector data type designed for storing embeddings. To enable semantic similarity queries, you must build a vector index on these columns. This allows you to run top-k searches efficiently, retrieving results based on the closeness of embeddings rather than simple keyword or numeric matches. Refer to Oracle� 23c documentation for best practices on vector data indexing and query optimization.
- A. Incorrect.
A. Incorrect. Storing embeddings as plain text and indexing with a text-based index will not enable true similarity (vector) searches. A text index is designed for keyword matching, not vector-based proximity queries.
- B. Incorrect.
B. Incorrect. Storing each dimension manually in a numeric column is cumbersome and does not leverage Oracle Database 23c� built-in vector capabilities. Numeric comparisons are slow for high-dimensional data, and you lose Oracle� optimized indexing for vectors.
- C. Correct.
C. Correct. Oracle Database 23c introduces a native vector data type and supports indexing for vectors, enabling efficient top-k similarity searches. This approach is specifically designed for storing embeddings and performing semantic queries in near real-time.
- D. Incorrect.
D. Incorrect. While using an external NoSQL store might work for some scenarios, it introduces additional complexity and latency. Periodically syncing large vectors back to Oracle Database 23c for queries disrupts near real-time performance and is less efficient compared to using the native vector type.