1Z0-184-25 Question 98
Single answerYou are designing a semantic search feature for a large e-commerce catalog on Oracle Cloud Infrastructure (OCI). Your team plans to generate high-dimensional vector embeddings for product descriptions, then store and query those embeddings efficiently in Oracle Database 23c. Which approach best facilitates scalable approximate nearest neighbor (ANN) queries on these embeddings while maintaining overall query performance?
- A
Create a specialized vector index on the embedding column in Oracle Database 23c and use the approximate similarity search operators for queries.
- B
Use a B-tree index on a single-column numeric representation (e.g., sum of all embedding values) to simplify indexing.
- C
Store the embeddings as JSON files in Oracle Object Storage and query them using Oracle Functions for any ANN operation.
- D
Convert the embeddings to text tokens and employ Oracle Database 23c full-text search on them for semantic matching.
Show answer and explanation
Correct answer: A
Explanation
Oracle Database 23c introduces powerful vector data type support and specialized vector indexes for approximate nearest neighbor (ANN) searches, which are essential for scalable semantic search on high-dimensional embeddings. By leveraging this native functionality, you can achieve more accurate results and better performance compared to traditional indexing methods. Refer to the Oracle Database 23c documentation and the 'Vector Search in Oracle Database' best-practices guide for details on creating and querying vector indexes to implement ANN search in your applications.
- A. Correct.
Option 1 is correct. Oracle Database 23c introduces native support for vector data types and specialized vector indexes. This enables efficient approximate nearest neighbor (ANN) queries on high-dimensional embeddings, making it the most scalable and performant approach for semantic search in an OCI environment.
- B. Incorrect.
Option 2 is incorrect. A B-tree index on a single numeric representation (like a summed embedding) discards much of the semantic information contained in high-dimensional vectors, which leads to poor recall and precision in ANN queries.
- C. Incorrect.
Option 3 is incorrect. Storing raw embeddings as JSON in Object Storage does not provide a built-in mechanism for efficient vector similarity searches. While you could implement custom logic with Oracle Functions, it would not scale as effectively as using specialized in-database indexing for ANN.
- D. Incorrect.
Option 4 is incorrect. Full-text search indexes are optimized for token-based searches, not numerical vectors. Converting embeddings to tokens would lose the continuous semantic relationships that vector-based approaches are designed to capture.