Question 540
A team responsible for a multilingual research-document retriever needs to represent passages as numeric vectors suitable for semantic-neighbor retrieval; the acceptance check is that each selected passage has an embedding of the expected dimensions. Which capability should be selected?
- Run a hybrid keyword and vector search
- Generate embeddings for indexed text
- Enable semantic ranking
- Run a vector nearest-neighbor query
Explanation
Correct Answer
B:An embedding model encodes text into vector coordinates that preserve useful semantic relationships. For this multilingual research-document retriever, the intended evidence is that each selected passage has an embedding of the expected dimensions. A keyword tokenizer alone does not create the numerical vector values.
Incorrect Answers
A:Hybrid search executes both text and vector retrieval and fuses result rankings into a shared result set; it may be used to combine lexical matching with semantic vector retrieval for better coverage in a multilingual research-document retriever, but the current requirement is instead to represent passages as numeric vectors suitable for semantic-neighbor retrieval. That separate capability does not prove that each selected passage has an embedding of the expected dimensions; A vector-only request does not directly include a BM25 lexical matching branch, so selecting that alternative would not meet the stated acceptance check.
C:Semantic ranker processes already retrieved text candidates and may extract captions and answer-like passages from them; it may be used to rerank an initial candidate set using language-model relevance and return supported captions or answers in a multilingual research-document retriever, but the current requirement is instead to represent passages as numeric vectors suitable for semantic-neighbor retrieval. That separate capability does not prove that each selected passage has an embedding of the expected dimensions; It does not create index entries or hallucinate new source facts as a generation model might, so selecting that alternative would not meet the stated acceptance check.
D:Vector search compares numeric representations according to the index's configured vector algorithms and search fields; it may be used to return index entries whose stored vectors are closest to the query vector in a multilingual research-document retriever, but the current requirement is instead to represent passages as numeric vectors suitable for semantic-neighbor retrieval. That separate capability does not prove that each selected passage has an embedding of the expected dimensions; It does not require a literal exact word overlap between query and result, so selecting that alternative would not meet the stated acceptance check.
Question 541
The receipt model returns a tax value with uncertain confidence. What should be done for reimbursement decisions?
- Use paragraphs, layout roles, and bounding regions from the Layout analysis
- Preserve the text reading order and page structure returned by the service
- Confirm the model's supported document types and regional coverage
- Apply a supported confidence threshold and manual verification for important fields
Explanation
Correct Answer
D:A recognized numeric value is not guaranteed correct simply because it is present. In the shared-services expense document processor, confirm the fix by retesting the affected behavior after applying a supported confidence threshold and manual verification for important fields.
Incorrect Answers
A:Although useful elsewhere in the shared-services expense document processor, page structure helps reconstruct content organization beyond arbitrary token order; the present case is different and calls for the team to apply a supported confidence threshold and manual verification for important fields.
B:Alphabetical sorting destroys source order; OCR results should be reconstructed using reading-order metadata in its own workflow, but that is a different correction from the present requirement to apply a supported confidence threshold and manual verification for important fields. The reported issue would remain.
C:That change addresses another failure mode: prebuilt ID extraction has bounded input support; unsupported formats may need another workflow; it does not address this case, for which the relevant intervention is to apply a supported confidence threshold and manual verification for important fields.
Question 542
The company sends unsupported national ID formats and assumes the prebuilt model handles every jurisdiction. What must be checked?
- Preserve the text reading order and page structure returned by the service
- Use paragraphs, layout roles, and bounding regions from the Layout analysis
- Validate missing or low-confidence invoice fields and route exceptions for review
- Confirm the model's supported document types and regional coverage
Explanation
Correct Answer
D:Prebuilt ID extraction has bounded input support; unsupported formats may need another workflow. In the shared-services expense document processor, confirm the fix by retesting the affected behavior after confirming the model's supported document types and regional coverage.
Incorrect Answers
A:Alphabetical sorting destroys source order; OCR results should be reconstructed using reading-order metadata in its own workflow, but that is a different correction from the present requirement to confirm the model's supported document types and regional coverage. The reported issue would remain.
B:Although useful elsewhere in the shared-services expense document processor, page structure helps reconstruct content organization beyond arbitrary token order; the present case is different and calls for the team to confirm the model's supported document types and regional coverage.
C:That change addresses another failure mode: prebuilt models are probabilistic; not every document exposes every recognized field reliably; it does not address this case, for which the relevant intervention is to confirm the model's supported document types and regional coverage.
Question 543
Accounts payable needs invoice number, supplier name, and due date, not an undifferentiated text transcript. Which output mapping helps?
- Preserve page identifiers and text spans or polygon coordinates
- Use Layout model table cells with row, column, and span metadata
- Use the prebuilt ID Document model and validate its typed identity fields
- Use the Invoice model's structured invoice fields and their confidence data
Explanation
Correct Answer
D:Prebuilt invoice extraction identifies domain-specific fields suitable for accounting workflows. In the shared-services expense document processor, confirm the fix by retesting the affected behavior after using the Invoice model's structured invoice fields and their confidence data.
Incorrect Answers
A:That change addresses another failure mode: read output includes location information on supported inputs, enabling verification against the source page; it does not address this case, for which the relevant intervention is to use the Invoice model's structured invoice fields and their confidence data.
B:Layout extraction preserves table structure that plain concatenated OCR text often loses in its own workflow, but that is a different correction from the present requirement to use the Invoice model's structured invoice fields and their confidence data. The reported issue would remain.
C:Although useful elsewhere in the shared-services expense document processor, the model extracts relevant fields from supported IDs; its output is data extraction, not independent proof of identity; the present case is different and calls for the team to use the Invoice model's structured invoice fields and their confidence data.
Question 544
In a shared-services expense document processor, engineers must obtain printed or handwritten text from scanned pages without specialized business-field extraction. The expected evidence is that the result contains page text and reading-order information. Which implementation is the best match?
- Use the prebuilt Read model
- Use the prebuilt Identity Document model
- Use the prebuilt Receipt model
- Use the prebuilt Invoice model
Explanation
Correct Answer
A:Document Intelligence Read focuses on OCR and text locations across supported document inputs. For this shared-services expense document processor, the intended evidence is that the result contains page text and reading-order information. It does not classify invoice totals into a dedicated invoice schema by default.
Incorrect Answers
B:The ID Document model returns structured fields from supported identity-document layouts; it may be used to extract recognized fields from supported passports or identification documents in a shared-services expense document processor, but the current requirement is instead to obtain printed or handwritten text from scanned pages without specialized business-field extraction. That separate capability does not prove that the result contains page text and reading-order information; It is not an identity authorization service and does not verify whether someone may access an Azure resource, so selecting that alternative would not meet the stated acceptance check.
C:The Receipt model is designed for common retail receipt fields without requiring a fresh custom labeled dataset; it may be used to recognize merchant, line item, tax, and total information from purchase receipts in a shared-services expense document processor, but the current requirement is instead to obtain printed or handwritten text from scanned pages without specialized business-field extraction. That separate capability does not prove that the result contains page text and reading-order information; An invoice-specific model is less directly aligned with point-of-sale slips, so selecting that alternative would not meet the stated acceptance check.
D:The Invoice model recognizes invoice-related fields such as vendor data, dates, and totals from supported formats; it may be used to extract supplier and amount fields from standard invoice documents in a shared-services expense document processor, but the current requirement is instead to obtain printed or handwritten text from scanned pages without specialized business-field extraction. That separate capability does not prove that the result contains page text and reading-order information; It does not expect to recognize the fields of an identity card instead, so selecting that alternative would not meet the stated acceptance check.
Question 545
A shared-services expense document processor must retrieve tables, paragraphs, and selection marks with document structure. Reviewers will accept the implementation when the output identifies table cells and page layout elements. Which approach best fulfills that requirement?
- Use the prebuilt Layout model
- Use the prebuilt Identity Document model
- Use the prebuilt Invoice model
- Use the prebuilt Receipt model
Explanation
Correct Answer
A:Document Intelligence Layout extracts structural elements of pages such as tables and selection marks where supported. For this shared-services expense document processor, the intended evidence is that the output identifies table cells and page layout elements. It is not chiefly a domain-specific purchase-invoice field classifier.
Incorrect Answers
B:The ID Document model returns structured fields from supported identity-document layouts; it may be used to extract recognized fields from supported passports or identification documents in a shared-services expense document processor, but the current requirement is instead to retrieve tables, paragraphs, and selection marks with document structure. That separate capability does not prove that the output identifies table cells and page layout elements; It is not an identity authorization service and does not verify whether someone may access an Azure resource, so selecting that alternative would not meet the stated acceptance check.
C:The Invoice model recognizes invoice-related fields such as vendor data, dates, and totals from supported formats; it may be used to extract supplier and amount fields from standard invoice documents in a shared-services expense document processor, but the current requirement is instead to retrieve tables, paragraphs, and selection marks with document structure. That separate capability does not prove that the output identifies table cells and page layout elements; It does not expect to recognize the fields of an identity card instead, so selecting that alternative would not meet the stated acceptance check.
D:The Receipt model is designed for common retail receipt fields without requiring a fresh custom labeled dataset; it may be used to recognize merchant, line item, tax, and total information from purchase receipts in a shared-services expense document processor, but the current requirement is instead to retrieve tables, paragraphs, and selection marks with document structure. That separate capability does not prove that the output identifies table cells and page layout elements; An invoice-specific model is less directly aligned with point-of-sale slips, so selecting that alternative would not meet the stated acceptance check.
Question 546
A cashier receipt lists merchant, tax, and total, but the app uses the invoice model and gets inconsistent field names. What should be selected?
- Use Layout model table cells with row, column, and span metadata
- Use the Invoice model's structured invoice fields and their confidence data
- Use the prebuilt ID Document model and validate its typed identity fields
- Use the Receipt model and map its receipt-specific output schema
Explanation
Correct Answer
D:Point-of-sale receipts have a different document structure from supplier invoices. In the shared-services expense document processor, confirm the fix by retesting the affected behavior after using the Receipt model and map its receipt-specific output schema.
Incorrect Answers
A:That change addresses another failure mode: layout extraction preserves table structure that plain concatenated OCR text often loses; it does not address this case, for which the relevant intervention is to use the Receipt model and map its receipt-specific output schema.
B:Although useful elsewhere in the shared-services expense document processor, prebuilt invoice extraction identifies domain-specific fields suitable for accounting workflows; the present case is different and calls for the team to use the Receipt model and map its receipt-specific output schema.
C:The model extracts relevant fields from supported IDs; its output is data extraction, not independent proof of identity in its own workflow, but that is a different correction from the present requirement to use the Receipt model and map its receipt-specific output schema. The reported issue would remain.
Question 547
An expense report must check whether the sum of receipt line items agrees with the total. What post-processing is appropriate?
- Consume supported selection-mark detection results
- Validate the extracted item amounts and total with application-side arithmetic
- Use a separate identity verification and authorization process before granting access
- Extract the recognized invoice items and their structured item fields
Explanation
Correct Answer
B:Document extraction returns recognized values; financial consistency checks remain an application responsibility. In the shared-services expense document processor, confirm the fix by retesting the affected behavior after validating the extracted item amounts and total with application-side arithmetic.
Incorrect Answers
A:Selection marks are visual document elements; their state is not reliably recoverable from text lines alone in its own workflow, but that is a different correction from the present requirement to validate the extracted item amounts and total with application-side arithmetic. The reported issue would remain.
C:That change addresses another failure mode: oCR or field extraction does not establish document authenticity or the user's right to an account; it does not address this case, for which the relevant intervention is to validate the extracted item amounts and total with application-side arithmetic.
D:Although useful elsewhere in the shared-services expense document processor, line items convey quantity, description, and amounts that a single total cannot represent; the present case is different and calls for the team to validate the extracted item amounts and total with application-side arithmetic.
Question 548
A team calls the Read model expecting it to output a purchase order's vendor-tax field by business name. What is the misunderstanding?
- Use a separate identity verification and authorization process before granting access
- Consume supported selection-mark detection results
- Extract the recognized invoice items and their structured item fields
- Use a field-specific prebuilt or custom extraction model for typed business fields
Explanation
Correct Answer
D:Read primarily returns text and its structure, not a universal named business schema. In the shared-services expense document processor, confirm the fix by retesting the affected behavior after using a field-specific prebuilt or custom extraction model for typed business fields.
Incorrect Answers
A:That change addresses another failure mode: oCR or field extraction does not establish document authenticity or the user's right to an account; it does not address this case, for which the relevant intervention is to use a field-specific prebuilt or custom extraction model for typed business fields.
B:Although useful elsewhere in the shared-services expense document processor, selection marks are visual document elements; their state is not reliably recoverable from text lines alone; the present case is different and calls for the team to use a field-specific prebuilt or custom extraction model for typed business fields.
C:Line items convey quantity, description, and amounts that a single total cannot represent in its own workflow, but that is a different correction from the present requirement to use a field-specific prebuilt or custom extraction model for typed business fields. The reported issue would remain.
Question 549
Recognized text becomes garbled after the app reorders OCR lines alphabetically. What should it use instead?
- Apply a supported confidence threshold and manual verification for important fields
- Validate missing or low-confidence invoice fields and route exceptions for review
- Use paragraphs, layout roles, and bounding regions from the Layout analysis
- Preserve the text reading order and page structure returned by the service
Explanation
Correct Answer
D:Alphabetical sorting destroys source order; OCR results should be reconstructed using reading-order metadata. In the shared-services expense document processor, confirm the fix by retesting the affected behavior after preserving the text reading order and page structure returned by the service.
Incorrect Answers
A:A recognized numeric value is not guaranteed correct simply because it is present in its own workflow, but that is a different correction from the present requirement to preserve the text reading order and page structure returned by the service. The reported issue would remain.
B:That change addresses another failure mode: prebuilt models are probabilistic; not every document exposes every recognized field reliably; it does not address this case, for which the relevant intervention is to preserve the text reading order and page structure returned by the service.
C:Although useful elsewhere in the shared-services expense document processor, page structure helps reconstruct content organization beyond arbitrary token order; the present case is different and calls for the team to preserve the text reading order and page structure returned by the service.
Question 550
An invoice includes a line-item table, and the system stores only the total. What field group should be mapped?
- Validate the extracted item amounts and total with application-side arithmetic
- Extract the recognized invoice items and their structured item fields
- Use a separate identity verification and authorization process before granting access
- Consume supported selection-mark detection results
Explanation
Correct Answer
B:Line items convey quantity, description, and amounts that a single total cannot represent. In the shared-services expense document processor, confirm the fix by retesting the affected behavior after extracting the recognized invoice items and their structured item fields.
Incorrect Answers
A:Although useful elsewhere in the shared-services expense document processor, document extraction returns recognized values; financial consistency checks remain an application responsibility; the present case is different and calls for the team to extract the recognized invoice items and their structured item fields.
C:That change addresses another failure mode: oCR or field extraction does not establish document authenticity or the user's right to an account; it does not address this case, for which the relevant intervention is to extract the recognized invoice items and their structured item fields.
D:Selection marks are visual document elements; their state is not reliably recoverable from text lines alone in its own workflow, but that is a different correction from the present requirement to extract the recognized invoice items and their structured item fields. The reported issue would remain.