1Z0-184-25 Question 109
Single answerYour company wants to implement a semantic search feature for a large collection of customer product reviews stored in an Oracle Autonomous Database. They plan to generate vector embeddings directly in the database so they can perform similarity searches (e.g., to find reviews most similar to a particular query). Which approach meets these requirements by leveraging native database functionality for both generating and managing those embeddings?
- A
Use Oracle Machine Learning to generate embeddings, store them in a dedicated VECTOR column type, and create a vector index for similarity search queries.
- B
Import customer reviews as text into Oracle Text indexes, which automatically generate vector embeddings for semantic search.
- C
Export the reviews to OCI Data Science, generate embeddings externally with a Python library, and store them as a BLOB in the database using a traditional B-tree index.
- D
Rely on a custom stored procedure in PL/SQL to compute embeddings and store them in multiple numeric columns without any special index.
Show answer and explanation
Correct answer: A
Explanation
In Oracle Database (including Autonomous Database deployments), you can use Oracle Machine Learning to generate vector embeddings for text data. By storing them in a specialized VECTOR column with a vector index, you can run neighbor and similarity searches natively. This approach ensures high performance and simplified data management. Reference: Oracle Database documentation on vector data types and Oracle Machine Learning highlights this recommended workflow for semantic search use cases.
- A. Correct.
Correct. Oracle supports generating embeddings through Oracle Machine Learning capabilities (for instance, using built-in or custom models). By storing these embeddings in a VECTOR column and creating a vector index, you can efficiently perform approximate nearest neighbor searches natively within the database.
- B. Incorrect.
Incorrect. Oracle Text provides full-text indexing and categorization functionality, but it does not automatically generate vector embeddings for semantic search. You still need a dedicated workflow or model to create vector embeddings.
- C. Incorrect.
Incorrect. Generating embeddings externally and storing them in a BLOB with a regular B-tree index can work, but it does not let you leverage the database� built-in vector index functionality for fast similarity search. An external generation step also complicates data management and consistency.
- D. Incorrect.
Incorrect. While you could store embeddings in manually managed numeric columns via PL/SQL, you would lose the benefits of Oracle� dedicated vector data type and specialized vector index, making similarity searches much less efficient.