1Z0-184-25 Question 95
Single answerYou are developing a semantic search feature for a content recommendation system on Oracle Cloud Infrastructure. Your team generates numeric vector embeddings for documents using OCI Data Science. You want to store these vector embeddings in a way that allows approximate nearest neighbor (ANN) queries to quickly find semantically similar documents. Which approach should you implement to achieve this while ensuring minimal overhead and efficient query performance?
- A
Store embeddings in a BLOB column in Oracle Database and use a standard B-tree index to perform ANN queries.
- B
Use the Vector data type introduced in Oracle Database 23c, store embeddings as vector columns, and create a specialized vector index for ANN queries.
- C
Convert embeddings into textual form, store them in a VARCHAR2 column, and rely on Oracle Text indexing for semantic similarity.
- D
Keep embeddings in Oracle Object Storage as text files and perform real-time ANN by referencing them through REST APIs.
Show answer and explanation
Correct answer: B
Explanation
For semantic search based on numeric embeddings, Oracle Database 23c offers a dedicated Vector data type and specialized vector indexing to enable efficient approximate nearest neighbor queries. By storing embeddings as vectors and creating a vector index, you can accelerate similarity-based searches at scale. Refer to the Oracle Database 23c documentation for details on how to create, store, and index vector columns for high-performance semantic queries in real-world applications.
- A. Incorrect.
Option 1 is incorrect because a standard B-tree index on BLOB data is not optimized for high-dimensional vector searches, and it would be inefficient for ANN queries.
- B. Correct.
Option 2 is correct. Oracle Database 23c introduces a Vector data type along with specialized indexing for vector columns, making it possible to efficiently run approximate nearest neighbor queries on numeric embeddings.
- C. Incorrect.
Option 3 is incorrect because Oracle Text is primarily designed for textual searches and does not offer native support for numeric vector operations. Converting numeric embeddings to text would be suboptimal and likely degrade performance.
- D. Incorrect.
Option 4 is incorrect because storing embeddings as text files in Oracle Object Storage and retrieving them via REST is not tailored for low-latency ANN queries. This approach would introduce unnecessary overhead and complicate query logic.