Explanation:
Option B is the correct answer because it best addresses all the requirements:
Handles long documents (5-100 pages): Amazon Bedrock Knowledge Bases are designed to handle large documents through intelligent chunking and retrieval mechanisms.
Preserves relationships between financial metrics and explanatory text: By segmenting documents based on structural layout, the solution maintains the contextual relationship between tables and their surrounding narrative text. This is crucial for financial documents where tables are often accompanied by explanatory commentary.
Provides contextually accurate responses: The RAG approach retrieves relevant chunks of information from the knowledge base, ensuring responses are grounded in the actual document content.
Includes citations: The solution explicitly mentions including citations that reference original source materials, which is important for financial analysis where traceability is critical.
Scalable retrieval: Knowledge Bases provide managed, scalable retrieval infrastructure that can handle multiple documents and queries efficiently.
Why other options are incorrect:
Option A: Processing entire documents as single inputs would be inefficient for 5-100 page documents and may exceed token limits. It also doesn't provide structured retrieval or citations.
Option C: Using fixed-size chunking with Lambda functions doesn't preserve structural relationships between tables and text, which is essential for financial documents.
Option D: Creating separate applications for structured and unstructured data would fragment the information and make it difficult to maintain relationships between tables and their explanatory text that naturally occur together in financial documents.
Ultimate access to all questions.
No comments yet.
A financial services company wants to develop an Amazon Bedrock application that gives analysts the ability to query quarterly earnings reports and financial statements. The financial documents are typically 5–100 pages long and contain both tabular data and text. The application must provide contextually accurate responses that preserve the relationship between financial metrics and their explanatory text. To support accurate and scalable retrieval, the application must incorporate document segmentation and context management strategies.
Which solution will meet these requirements?
A
Use a direct model invocation approach that uses Anthropic Claude to process each financial document as a single input. Use fine-tuned prompts that instruct the model to parse tables and text separately.
B
Use Amazon Bedrock Knowledge Bases to create a Retrieval Augmented Generation (RAG) application that retrieves relevant information from contextually chunked sections of financial documents. Segment documents based on their structural layout. Include citations that reference the original source materials.
C
Deploy an Amazon Bedrock agent that has an action group that calls custom AWS Lambda functions to analyze financial documents. Configure the Lambda functions to perform fixed-size chunking when a user submits a query about financial metrics.
D
Create one specialized Amazon Bedrock application that is optimized for structured data. Create a second application that is optimized for unstructured data. Configure each application to use a tailored chunking strategy that is suited to the application's content type. Implement logic to link queries to the appropriate sources.