Databricks Generative AI Engineer Associate Question 51
Select 3You are building a Retrieval-Augmented Generation (RAG) application using Databricks and need to filter extraneous content from source documents to improve the quality of responses. Which methods are appropriate to ensure the quality of the source documents used for retrieval?
- A
Removing irrelevant sections of documents based on business context
- B
Filtering out duplicate documents within the source repository
- C
Including all content, even if unrelated, to maximize the information available
- D
Using metadata tags to identify and exclude low-quality or outdated documents
- E
Applying a summarization model to reduce document size before indexing
Show answer and explanation
Correct answers: A, B, D
Explanation
Filtering extraneous content is essential for maintaining the quality of a RAG application. Methods like removing irrelevant sections, deduplicating documents, and leveraging metadata for filtering ensure that only high-quality and relevant information is retained in the source documents. These practices help improve the retrieval process and, consequently, the overall quality of generated responses.
- A. Correct.
Correct. Removing irrelevant sections ensures that only content aligned with the application's purpose is included, improving retrieval quality.
- B. Correct.
Correct. Duplicates within the source repository can degrade the retrieval process by introducing redundancy and noise.
- C. Incorrect.
Incorrect. Including unrelated content increases the risk of generating low-quality responses, as the retrieval step may pull irrelevant information.
- D. Correct.
Correct. Metadata-based filtering helps exclude documents that are outdated or of low quality, leading to better retrieval outcomes.
- E. Incorrect.
Incorrect. While summarization can reduce document size, it may inadvertently omit details critical to the RAG application, so it is not a reliable filtering method.