Filter Extraneous Content in Source Documents That Degrades Quality of a RAG Application GENERATIVE-AI-ENGINEER-ASSOCIATE Practice Questions (Page 3)

Questions 11–15

1. 11application · medium

   A tech company is building a RAG application over its internal code documentation. The documentation includes code snippets, API references, and explanatory text. The team notices that the code snippets are often duplicated across multiple pages, and this duplication is causing retrieval to return redundant chunks. They want to remove the duplicate code snippets while preserving the explanatory text. Which approach is most appropriate?

   • AUse a regex pattern to remove any text that contains curly braces, as these are likely code snippets.
   • BRemove all lines of code from the documentation, keeping only the explanatory text.
   • CIncrease the chunk size so that each chunk contains more explanatory text, diluting the impact of the duplicate code snippets.
   • DUse a near-duplicate detection algorithm (e.g., MinHash) on the code snippets to identify and remove near-identical copies, keeping only one canonical version.
   Reveal Solution Discussion Select an answer first
2. 12application · medium

   A customer support team is building a RAG application over internal troubleshooting guides. The guides contain step-by-step instructions, but also include 'Notes', 'Warnings', and 'Tips' boxes that are visually distinct in the original PDF but are extracted as plain text. These boxes often contain critical caveats (e.g., 'Do not run this command on a production server'). The team applies a filter that removes all text between 'Note:' and the next blank line to clean up the guides. After deployment, they find that the RAG system now suggests unsafe commands. What is the most likely cause?

   • AThe filter removed the 'Warnings' and 'Tips' boxes that contained the safety caveats, leaving only the main instructions.
   • BThe filter was too conservative and failed to remove the 'Notes' boxes, leaving them in the index and causing confusion.
   • CThe filter was applied after chunking, so the embeddings were computed on the noisy text, and the filter had no effect.
   • DThe retrieval system is not configured to rank safety warnings higher than other text.
   Reveal Solution Discussion Select an answer first
4. 13foundation · easy

   A RAG pipeline ingests HTML pages that contain navigation menus and footer links. Which filtering technique is MOST appropriate to remove these elements?

   • AAn ML-based classifier trained to identify boilerplate content
   • BA manual review of every page to delete unwanted sections
   • CA regex pattern that matches common navigation link text
   • DA heuristic that removes content inside <nav> and <footer> tags
   Reveal Solution Discussion Select an answer first
5. 14foundation · easy

   Which practice BEST preserves relevant information while filtering extraneous content?

   • AFiltering out all sentences shorter than 10 words
   • BApplying the same filter to all document types without review
   • CRemoving all content that appears in more than one document
   • DUsing a conservative filter that targets only clearly non-informative patterns
   Reveal Solution Discussion Select an answer first
6. 15application · medium

   A government agency is building a RAG application over public consultation documents. These documents contain a main proposal text, followed by a long list of 'Frequently Asked Questions' (FAQs) and a 'Glossary of Terms'. The FAQs and glossary are repeated in every document. The team wants to filter these to improve retrieval. Which approach is most effective?

   • AUse a text summarization model to condense the entire document, removing the FAQ and Glossary sections as part of the summarization.
   • BIdentify and remove the FAQ and Glossary sections using document structure markers (e.g., '## FAQ', '## Glossary') and regex to match the section boundaries.
   • CRemove any sentence that contains a question mark, as these are likely part of the FAQ section.
   • DRemove any sentence that contains a term defined in the glossary.
   Reveal Solution Discussion Select an answer first