1Z0-184-25 Question 110
Single answerYou have a large catalog of text documents in your Oracle Database, and you want to enable semantic search by generating and storing vector embeddings for each document within the same database environment. Your objective is to minimize data movement while leveraging built-in Oracle capabilities. Which approach should you use to generate the embeddings directly inside the database and store them for efficient similarity queries?
- A
A. Export the documents to an external Python environment, generate embeddings with a third-party library, and re-import them into Oracle Database using Data Pump.
- B
B. Use Oracle Machine Learning with Python (OML4Py) to run an in-database script that applies a pretrained embedding model to the stored documents and saves the results in a vector column.
- C
C. Create a B-Tree index on the text column to directly store and retrieve embeddings without additional modifications.
- D
D. Use Oracle Text to automatically generate and store vector embeddings in a hidden metadata table.
Show answer and explanation
Correct answer: B
Explanation
Oracle Machine Learning with Python (OML4Py) provides in-database execution of Python scripts, allowing data scientists and developers to run machine learning or embedding-generation workflows directly within Oracle Database. By storing vector embeddings in a dedicated vector column and creating specialized vector indexes (introduced in modern releases of Oracle Database), you can enable fast semantic or similarity-based searches on textual data. Refer to Oracle Machine Learning documentation and Oracle Database documentation on vector data types for detailed guidance.
- A. Incorrect.
A. Incorrect. Exporting documents to an external environment and re-importing embeddings increases data movement, complexity, and operational overhead. While possible, it does not leverage in-database processing.
- B. Correct.
B. Correct. Oracle Machine Learning with Python (OML4Py) allows you to run Python scripts natively inside the database. You can integrate a pretrained embedding model, generate embeddings for each document, and store them in a vector column�enabling efficient similarity queries without leaving the database environment.
- C. Incorrect.
C. Incorrect. B-Tree indexes are not designed for vector data or similarity searches. You need a dedicated vector data type and index (introduced in newer versions of Oracle Database) for efficient similarity-based operations.
- D. Incorrect.
D. Incorrect. Oracle Text provides sophisticated text indexing and search, but it does not automatically generate vector embeddings. Oracle Text is keyword-based, rather than embedding-based, and uses different mechanisms for indexing and searching textual content.