1Z0-184-25 Question 9
Select 2Your team is building a recommendation engine on Oracle Autonomous Database in OCI to serve personalized product suggestions. They plan to store text embeddings and run semantic queries (e.g., finding items with similar descriptions) using those embeddings. Which two steps should you implement to ensure efficient similarity searches using the Vector data type?
- A
Store embeddings in a BLOB column with a standard B-tree index for textual queries.
- B
Use the dedicated Vector data type, specifying the dimension of each embedding, so that specialized vector search algorithms can be leveraged.
- C
Load embeddings as a JSON array and rely on built-in full-text indexes for semantic retrieval.
- D
Create an approximate nearest neighbor (ANN) index on the Vector column to accelerate similarity queries.
- E
Use a VARCOLL data type with default indexing, relying on the database to infer semantic similarity.
Show answer and explanation
Correct answers: B, D
Explanation
When working with embeddings in Oracle Autonomous Database, you should use the new Vector data type (dimension specified) and create an approximate nearest neighbor (ANN) index to support efficient semantic similarity searches. Oracle Database 23c introduced native capabilities for vector storage and querying, and these capabilities are included or enhanced in future Autonomous Database versions on Oracle Cloud Infrastructure. Refer to the official Oracle Database documentation on Vector Data Types and ANN Indexes for details on configuration and best practices.
- A. Incorrect.
Option 1 is incorrect. Storing embeddings as BLOBs with a standard B-tree index does not allow efficient similarity or semantic matching, because B-trees are designed for exact or range matching, not vector-based similarity computations.
- B. Correct.
Option 2 is correct. Oracle Database supports a specialized Vector data type (introduced in Oracle Database 23c and available in future Autonomous Database releases) that allows specifying the dimensionality of embeddings. This is essential for subsequent vector-based similarity and semantic queries.
- C. Incorrect.
Option 3 is incorrect. Storing embeddings in JSON alone does not provide native vector search capabilities. Full-text indexes are optimized for text-based keyword searches, not numeric vector similarity comparisons.
- D. Correct.
Option 4 is correct. After defining a Vector column, creating an approximate nearest neighbor (ANN) index is crucial for performing high-performance semantic similarity searches at scale. It optimizes distance-based lookups among high-dimensional vectors.
- E. Incorrect.
Option 5 is incorrect. Relying on a generic structure like VARCOLL with default indexes does not provide robust vector search capabilities. Vector data needs specialized indexing to handle similarity queries efficiently.