Topic 02 Practice Test 3 covers Multimodal Input Validation, Processing, Formatting, and Data Quality in Domain 1 (Task 1.3). For broader exam preparation, review the AIP-C01 Exam Dumps. Every option includes focused technical reasoning tied to the tested decision.
Question 1
An application team is standardizing data before foundation-model inference. A support dataset contains a priority field that must be one of LOW, MEDIUM, HIGH, or CRITICAL before prompts are constructed. Which implementation best satisfies the requirement? Choose ONE.
- cross-field consistency validation
- AWS Glue Data Quality allowed-value rule
- S3 quarantine prefix for failed validation
- AWS Glue Data Quality uniqueness rule
Correct Answer: B
Correct Answer
Answer B is correct because AWS Glue Data Quality allowed-value rule is appropriate because it will validate a categorical column against an approved domain before downstream processing. The requirement centers on a situation where only enumerated labels are valid for a field consumed by prompting logic.
Incorrect Answers
Answer A is incorrect because cross-field consistency validation belongs in workflows where data can be individually well-typed yet internally inconsistent because it can evaluate related fields together so logically impossible combinations are rejected. This question instead asks for only enumerated labels are valid for a field consumed by prompting logic.
Answer C is incorrect because S3 quarantine prefix for failed validation targets a different issue by helping to isolate records that fail validation from the clean processing path while retaining them for investigation. It fits where invalid objects must be retained but prevented from reaching FM consumption. The deciding condition is only enumerated labels are valid for a field consumed by prompting logic.
Answer D is incorrect because AWS Glue Data Quality uniqueness rule can measure duplicate keys in a managed ruleset and fail the quality check when uniqueness drops. It fits when duplicate business keys in a tabular batch would cause repeated evidence. This scenario instead requires only enumerated labels are valid for a field consumed by prompting logic, so it targets a different problem.
Question 2
A platform team is reviewing a production GenAI data pipeline. An API sends JSON records to a GenAI service. A request must be rejected when customerId is missing or amount is sent as a string. Which implementation best satisfies the requirement? Choose ONE.
- Lambda JSON Schema validation
- AWS Glue Data Quality uniqueness rule
- S3 quarantine prefix for failed validation
- model input-size precheck
Correct Answer: A
Correct Answer
Answer A is correct because Select Lambda JSON Schema validation. Its purpose is to enforce the expected structure and data types for each event before invoking the model. The operational requirement is that small event payloads need custom application-level validation at ingestion time.
Incorrect Answers
Answer B is incorrect because AWS Glue Data Quality uniqueness rule can measure duplicate keys in a managed ruleset and fail the quality check when uniqueness drops, which matters when duplicate business keys in a tabular batch would cause repeated evidence. The present case is driven by small event payloads need custom application-level validation at ingestion time.
Answer C is incorrect because S3 quarantine prefix for failed validation belongs in workflows where invalid objects must be retained but prevented from reaching FM consumption because it can isolate records that fail validation from the clean processing path while retaining them for investigation. This question instead asks for small event payloads need custom application-level validation at ingestion time.
Answer D is incorrect because model input-size precheck is useful when requests exceeding model limits should be handled deterministically before calling the FM and can estimate payload size or token count before invocation and reject or route oversized input. The described workload requires small event payloads need custom application-level validation at ingestion time.
Question 3
A cloud engineering team must correct a data-processing weakness. A compliance workflow must keep malformed source documents for audit, yet the FM ingestion job must process only records that passed validation. Which implementation best satisfies the requirement? Choose ONE.
- required-field freshness validation
- S3 quarantine prefix for failed validation
- SageMaker Data Wrangler profiling
- checksum integrity verification
Correct Answer: B
Correct Answer
Answer B is correct because S3 quarantine prefix for failed validation is the direct remedy. It can isolate records that fail validation from the clean processing path while retaining them for investigation. The critical scenario condition is that invalid objects must be retained but prevented from reaching FM consumption.
Incorrect Answers
Answer A is incorrect because required-field freshness validation solves a different input problem. It helps to compare a source timestamp with an allowed staleness window before accepting the record, normally when the FM must operate only on sufficiently recent operational data. The current workflow needs invalid objects must be retained but prevented from reaching FM consumption.
Answer C is incorrect because SageMaker Data Wrangler profiling would be sound if the goal were a team must understand data quality problems before finalizing preprocessing logic; it can profile distributions, missingness, and anomalies interactively before defining production transformations. The requirement here is invalid objects must be retained but prevented from reaching FM consumption.
Answer D is incorrect because checksum integrity verification can compare a supplied cryptographic checksum with the downloaded object before preprocessing, which matters when large artifacts can be corrupted in transit or during transfer and must be verified before use. The present case is driven by invalid objects must be retained but prevented from reaching FM consumption.
Question 4
A regulated workflow requires deterministic preprocessing before inference. A record can be status=CLOSED only when closed_at is populated. The FM must not receive records that violate that business rule. Which implementation best satisfies the requirement? Choose ONE.
- AWS Glue Data Quality completeness rule
- cross-field consistency validation
- required-field freshness validation
- Lambda JSON Schema validation
Correct Answer: B
Correct Answer
Answer B is correct because cross-field consistency validation is the best fit because it can evaluate related fields together so logically impossible combinations are rejected. The workflow needs a case where data can be individually well-typed yet internally inconsistent.
Incorrect Answers
Answer A is incorrect because AWS Glue Data Quality completeness rule helps when a batch table has mandatory fields and incomplete rows must be rejected by allowing the pipeline to evaluate required columns for nulls before records enter the FM pipeline. The described issue is data can be individually well-typed yet internally inconsistent.
Answer C is incorrect because required-field freshness validation solves a different input problem. It helps to compare a source timestamp with an allowed staleness window before accepting the record, normally when the FM must operate only on sufficiently recent operational data. The current workflow needs data can be individually well-typed yet internally inconsistent.
Answer D is incorrect because Lambda JSON Schema validation is appropriate where small event payloads need custom application-level validation at ingestion time. It achieves that by helping to enforce the expected structure and data types for each event before invoking the model. This scenario instead requires data can be individually well-typed yet internally inconsistent.
Question 5
A developer is hardening an FM-consumption workflow. A recommendation workflow must not use inventory snapshots older than 10 minutes even if every field is otherwise valid. Which implementation best satisfies the requirement? Choose ONE.
- AWS Glue Data Quality numeric-range rule
- CloudWatch invalid-record-rate metric
- SageMaker Data Wrangler profiling
- required-field freshness validation
Correct Answer: D
Correct Answer
Answer D is correct because required-field freshness validation is appropriate because it will compare a source timestamp with an allowed staleness window before accepting the record. The requirement centers on a situation where the FM must operate only on sufficiently recent operational data.
Incorrect Answers
Answer A is incorrect because AWS Glue Data Quality numeric-range rule is designed to check numeric values against expected minimum and maximum boundaries as part of batch quality validation. That makes sense when out-of-range measurements would mislead the FM. The question is centered on the FM must operate only on sufficiently recent operational data.
Answer B is incorrect because CloudWatch invalid-record-rate metric can publish a quality metric and alarm when the proportion of rejected inputs exceeds an operational threshold. Its best fit is where operators need trend visibility into data-quality failures rather than only per-record rejection. The current case instead needs the FM must operate only on sufficiently recent operational data.
Answer C is incorrect because SageMaker Data Wrangler profiling helps when a team must understand data quality problems before finalizing preprocessing logic by allowing the pipeline to profile distributions, missingness, and anomalies interactively before defining production transformations. The described issue is the FM must operate only on sufficiently recent operational data.
Question 6
An application team is standardizing data before foundation-model inference. An application must compare a product photo with a written packaging policy and explain any visible mismatch. Which implementation best satisfies the requirement? Choose ONE.
- video key-frame extraction plus transcript pipeline
- event-driven S3 Lambda preprocessing
- language-aware transcription routing
- multimodal Bedrock request with image and text
Correct Answer: D
Correct Answer
Answer D is correct because Select multimodal Bedrock request with image and text. Its purpose is to send supported image content together with textual instructions to a multimodal foundation model. The operational requirement is that the task requires reasoning over visual content, not merely OCR text.
Incorrect Answers
Answer A is incorrect because video key-frame extraction plus transcript pipeline works for cases where full video is too large for direct use but both visual and spoken evidence matter by helping to reduce a video to representative visual frames and speech text before multimodal reasoning. The present scenario needs the task requires reasoning over visual content, not merely OCR text.
Answer B is incorrect because event-driven S3 Lambda preprocessing targets a neighboring concern. It can trigger lightweight preprocessing automatically when a new object lands in S3 when small independent objects should begin processing immediately without a polling batch job. The key requirement here is the task requires reasoning over visual content, not merely OCR text.
Answer C is incorrect because language-aware transcription routing is designed to detect or configure the spoken language so the speech service uses the appropriate recognition path. That makes sense when audio arrives in multiple languages and a single hard-coded language causes poor transcripts. The question is centered on the task requires reasoning over visual content, not merely OCR text.
Question 7
A platform team is reviewing a production GenAI data pipeline. Mobile uploads include very large TIFF images, but the multimodal request path accepts smaller JPEG or PNG images. Which implementation best satisfies the requirement? Choose ONE.
- SageMaker Processing batch media preprocessing
- parallel modality preprocessing
- image dimension and format normalization
- multimodal Bedrock request with image and text
Correct Answer: C
Correct Answer
Answer C is correct because image dimension and format normalization is the direct remedy. It can resize oversized images and convert them to a supported format before model invocation. The critical scenario condition is that input images exceed model/service constraints even though their visual content is valid.
Incorrect Answers
Answer A is incorrect because SageMaker Processing batch media preprocessing can help a pipeline run scalable, repeatable preprocessing jobs over large media batches without tying work to request-time compute, especially when thousands of files need scheduled transformation before FM consumption. The present requirement is input images exceed model/service constraints even though their visual content is valid, so it is not the best answer.
Answer B is incorrect because parallel modality preprocessing would improve another stage by enabling it to process independent audio, image, and text components concurrently before joining them for inference. It fits when multiple independent modalities create unnecessary latency when handled serially. The stated problem is input images exceed model/service constraints even though their visual content is valid.
Answer D is incorrect because multimodal Bedrock request with image and text targets a neighboring concern. It can send supported image content together with textual instructions to a multimodal foundation model when the task requires reasoning over visual content, not merely OCR text. The key requirement here is input images exceed model/service constraints even though their visual content is valid.
Question 8
A cloud engineering team must correct a data-processing weakness. A document job must run OCR, validate extracted fields, normalize text, and only then call the FM, with retry rules for each stage. Which implementation best satisfies the requirement? Choose ONE.
- tabular rows serialized as structured JSON
- Step Functions multimodal orchestration
- image dimension and format normalization
- parallel modality preprocessing
Correct Answer: B
Correct Answer
Answer B is correct because Step Functions multimodal orchestration is the best fit because it can coordinate separate extraction, normalization, validation, and FM steps with explicit retries and failure branches. The workflow needs a case where a multi-stage preprocessing workflow needs durable state and error handling.
Incorrect Answers
Answer A is incorrect because Using tabular rows serialized as structured JSON would let the team convert selected columns into explicit key-value objects so the FM receives unambiguous field names and values. That is appropriate when a tabular record must be represented clearly for inference rather than as an unlabeled value sequence. The scenario instead calls for a multi-stage preprocessing workflow needs durable state and error handling.
Answer C is incorrect because image dimension and format normalization can help a pipeline resize oversized images and convert them to a supported format before model invocation, especially when input images exceed model/service constraints even though their visual content is valid. The present requirement is a multi-stage preprocessing workflow needs durable state and error handling, so it is not the best answer.
Answer D is incorrect because parallel modality preprocessing would improve another stage by enabling it to process independent audio, image, and text components concurrently before joining them for inference. It fits when multiple independent modalities create unnecessary latency when handled serially. The stated problem is a multi-stage preprocessing workflow needs durable state and error handling.
Question 9
A regulated workflow requires deterministic preprocessing before inference. Scanned handwritten forms sometimes produce very low OCR confidence. The business requires verified text before the FM drafts a decision. Which implementation best satisfies the requirement? Choose ONE.
- low-confidence OCR routing to human review
- Amazon Textract document extraction
- SageMaker Processing batch media preprocessing
- multimodal Bedrock request with image and text
Correct Answer: A
Correct Answer
Answer A is correct because low-confidence OCR routing to human review is appropriate because it will branch documents with unreliable extraction scores to manual verification before FM use. The requirement centers on a situation where uncertain extracted text must not silently become authoritative model context.
Incorrect Answers
Answer B is incorrect because Amazon Textract document extraction addresses another concern by helping to extract printed text, forms, and table structure from scanned documents before downstream language processing. It fits where scanned forms need structured text and key-value information before an FM can reason over them. Here the deciding need is uncertain extracted text must not silently become authoritative model context.
Answer C is incorrect because Using SageMaker Processing batch media preprocessing would let the team run scalable, repeatable preprocessing jobs over large media batches without tying work to request-time compute. That is appropriate when thousands of files need scheduled transformation before FM consumption. The scenario instead calls for uncertain extracted text must not silently become authoritative model context.
Answer D is incorrect because multimodal Bedrock request with image and text has value because it can send supported image content together with textual instructions to a multimodal foundation model. Its proper fit is when the task requires reasoning over visual content, not merely OCR text. This question requires uncertain extracted text must not silently become authoritative model context.
Question 10
A developer is hardening an FM-consumption workflow. A financial PDF has dense tables where values such as ‘14.2’ are meaningless unless the metric and period headers stay associated. Which implementation best satisfies the requirement? Choose ONE.
- parallel modality preprocessing
- low-confidence OCR routing to human review
- table-header preservation during extraction
- large-file chunked preprocessing
Correct Answer: C
Correct Answer
Answer C is correct because Select table-header preservation during extraction. Its purpose is to retain column headers with row values when converting tables to text or structured records. The operational requirement is that separating values from their headers would destroy meaning needed by the FM.
Incorrect Answers
Answer A is incorrect because parallel modality preprocessing targets a different issue by helping to process independent audio, image, and text components concurrently before joining them for inference. It fits where multiple independent modalities create unnecessary latency when handled serially. The deciding condition is separating values from their headers would destroy meaning needed by the FM.
Answer B is incorrect because low-confidence OCR routing to human review can branch documents with unreliable extraction scores to manual verification before FM use. It fits when uncertain extracted text must not silently become authoritative model context. This scenario instead requires separating values from their headers would destroy meaning needed by the FM, so it targets a different problem.
Answer D is incorrect because large-file chunked preprocessing addresses another concern by helping to split a very large source artifact into bounded processing units before normalization and later reassembly. It fits where a single preprocessing job would exceed memory or request limits. Here the deciding need is separating values from their headers would destroy meaning needed by the FM.
Question 11
An application team is standardizing data before foundation-model inference. A support bot has a permanent instruction to answer only from company policy, plus a new customer question on every turn. Which implementation best satisfies the requirement? Choose ONE.
- speaker-label mapping to conversation roles
- base64 or binary image content with correct media type
- conversation attachment content blocks
- system instruction separated from user content
Correct Answer: D
Correct Answer
Answer D is correct because system instruction separated from user content is the direct remedy. It can place persistent behavioral instructions in the system field while keeping the user’s actual request in the user message. The critical scenario condition is that the application must clearly distinguish global behavior from per-turn input.
Incorrect Answers
Answer A is incorrect because speaker-label mapping to conversation roles belongs in workflows where source-system speaker names are not directly valid message roles for the inference API because it can map source labels such as AGENT and CUSTOMER into the model API’s supported conversational roles. This question instead asks for the application must clearly distinguish global behavior from per-turn input.
Answer B is incorrect because base64 or binary image content with correct media type is useful when visual input is valid but the request currently treats binary bytes as ordinary text and can encode image content using the request format and MIME type expected by the multimodal API. The described workload requires the application must clearly distinguish global behavior from per-turn input.
Answer C is incorrect because conversation attachment content blocks targets a different issue by helping to represent text and supported attachments as separate typed content blocks within one conversational message. It fits where a message combines written instructions with a document or image and each part must retain its type. The deciding condition is the application must clearly distinguish global behavior from per-turn input.
Question 12
A platform team is reviewing a production GenAI data pipeline. Conversation history is stored as timestamped records in DynamoDB. The request builder must reconstruct the exact user/assistant sequence. Which implementation best satisfies the requirement? Choose ONE.
- correct HTTP Content-Type for JSON inference
- ordered conversation-turn serialization
- provider-versioned request adapter
- Bedrock Converse message format
Correct Answer: B
Correct Answer
Answer B is correct because ordered conversation-turn serialization is the best fit because it can preserve chronological role order when converting stored chat history into the model request. The workflow needs a case where reordered dialog turns would change the meaning of follow-up questions.
Incorrect Answers
Answer A is incorrect because correct HTTP Content-Type for JSON inference would be sound if the goal were a syntactically valid JSON body is rejected because the protocol metadata is wrong; it can send the media type expected by the inference endpoint together with the serialized request body. The requirement here is reordered dialog turns would change the meaning of follow-up questions.
Answer C is incorrect because provider-versioned request adapter belongs in workflows where an application supports multiple model families whose native schemas evolve independently because it can route each selected model through an adapter that owns that model family’s request and response mapping. This question instead asks for reordered dialog turns would change the meaning of follow-up questions.
Answer D is incorrect because Bedrock Converse message format is useful when a multi-turn application needs a consistent conversational request format across supported models and can represent dialog turns with the standardized role/content message structure used by the Converse API. The described workload requires reordered dialog turns would change the meaning of follow-up questions.
Question 13
A cloud engineering team must correct a data-processing weakness. A customer comment contains quotes and newline characters. Hand-built string concatenation intermittently creates invalid Bedrock request JSON. Which implementation best satisfies the requirement? Choose ONE.
- system instruction separated from user content
- batch-inference record envelope
- model-specific Bedrock request serialization
- JSON escaping of untrusted text fields
Correct Answer: D
Correct Answer
Answer D is correct because JSON escaping of untrusted text fields is appropriate because it will use a serializer that escapes quotes, backslashes, and control characters instead of string concatenation. The requirement centers on a situation where raw user text can otherwise produce malformed JSON request bodies.
Incorrect Answers
Answer A is incorrect because system instruction separated from user content solves a different input problem. It helps to place persistent behavioral instructions in the system field while keeping the user’s actual request in the user message, normally when the application must clearly distinguish global behavior from per-turn input. The current workflow needs raw user text can otherwise produce malformed JSON request bodies.
Answer B is incorrect because batch-inference record envelope is appropriate where offline inference requires per-record association so outputs can be mapped back to source items. It achieves that by helping to emit each batch item using the service’s required record identifier and model-input envelope. This scenario instead requires raw user text can otherwise produce malformed JSON request bodies.
Answer C is incorrect because model-specific Bedrock request serialization would be sound if the goal were native InvokeModel payloads differ among model providers; it can build the JSON body expected by the selected model instead of assuming every provider uses the same native schema. The requirement here is raw user text can otherwise produce malformed JSON request bodies.
Question 14
A regulated workflow requires deterministic preprocessing before inference. A user message includes a short question and a product image. The request builder currently flattens both into one text string. Which implementation best satisfies the requirement? Choose ONE.
- provider-versioned request adapter
- system instruction separated from user content
- conversation attachment content blocks
- correct HTTP Content-Type for JSON inference
Correct Answer: C
Correct Answer
Answer C is correct because Select conversation attachment content blocks. Its purpose is to represent text and supported attachments as separate typed content blocks within one conversational message. The operational requirement is that a message combines written instructions with a document or image and each part must retain its type.
Incorrect Answers
Answer A is incorrect because provider-versioned request adapter helps when an application supports multiple model families whose native schemas evolve independently by allowing the pipeline to route each selected model through an adapter that owns that model family’s request and response mapping. The described issue is a message combines written instructions with a document or image and each part must retain its type.
Answer B is incorrect because system instruction separated from user content is relevant, but it is intended to place persistent behavioral instructions in the system field while keeping the user’s actual request in the user message when the application must clearly distinguish global behavior from per-turn input. The actual requirement is a message combines written instructions with a document or image and each part must retain its type.
Answer D is incorrect because correct HTTP Content-Type for JSON inference is appropriate where a syntactically valid JSON body is rejected because the protocol metadata is wrong. It achieves that by helping to send the media type expected by the inference endpoint together with the serialized request body. This scenario instead requires a message combines written instructions with a document or image and each part must retain its type.
Question 15
A developer is hardening an FM-consumption workflow. A contact-center transcript stores each turn as AGENT or CUSTOMER, but the model request requires standardized role values. Which implementation best satisfies the requirement? Choose ONE.
- base64 or binary image content with correct media type
- provider-versioned request adapter
- speaker-label mapping to conversation roles
- model-specific Bedrock request serialization
Correct Answer: C
Correct Answer
Answer C is correct because speaker-label mapping to conversation roles is the direct remedy. It can map source labels such as AGENT and CUSTOMER into the model API’s supported conversational roles. The critical scenario condition is that source-system speaker names are not directly valid message roles for the inference API.
Incorrect Answers
Answer A is incorrect because base64 or binary image content with correct media type is designed to encode image content using the request format and MIME type expected by the multimodal API. That makes sense when visual input is valid but the request currently treats binary bytes as ordinary text. The question is centered on source-system speaker names are not directly valid message roles for the inference API.
Answer B is incorrect because provider-versioned request adapter can route each selected model through an adapter that owns that model family’s request and response mapping. Its best fit is where an application supports multiple model families whose native schemas evolve independently. The current case instead needs source-system speaker names are not directly valid message roles for the inference API.
Answer D is incorrect because model-specific Bedrock request serialization is relevant, but it is intended to build the JSON body expected by the selected model instead of assuming every provider uses the same native schema when native InvokeModel payloads differ among model providers. The actual requirement is source-system speaker names are not directly valid message roles for the inference API.
Question 16
An application team is standardizing data before foundation-model inference. Copied web text contains runs of spaces and inconsistent CRLF/LF line endings that inflate prompts but do not add information. Which implementation best satisfies the requirement? Choose ONE.
- whitespace normalization
- exact-content deduplication
- Bedrock-assisted text reformatting with constrained output
- unit normalization with original-value retention
Correct Answer: A
Correct Answer
Answer A is correct because whitespace normalization is the best fit because it can collapse accidental repeated spaces and normalize line endings while preserving semantic paragraph boundaries. The workflow needs a case where formatting noise creates unnecessary tokens without carrying meaning.
Incorrect Answers
Answer B is incorrect because exact-content deduplication targets a neighboring concern. It can remove repeated identical text blocks before assembling the model context when duplicated passages waste tokens and can overweight one source. The key requirement here is formatting noise creates unnecessary tokens without carrying meaning.
Answer C is incorrect because Bedrock-assisted text reformatting with constrained output is designed to use an FM preprocessing step to transform irregular prose into a defined structure while validating the result afterward. That makes sense when source text is semistructured and deterministic parsing alone cannot reliably produce the required representation. The question is centered on formatting noise creates unnecessary tokens without carrying meaning.
Answer D is incorrect because unit normalization with original-value retention can convert measurements to a common unit while preserving the source value for traceability. Its best fit is where mixed units make comparison unreliable unless values are standardized. The current case instead needs formatting noise creates unnecessary tokens without carrying meaning.
Question 17
A platform team is reviewing a production GenAI data pipeline. A crawler stores the same policy paragraph in navigation, mobile, and desktop versions, causing it to appear three times in the prompt context. Which implementation best satisfies the requirement? Choose ONE.
- date-format canonicalization
- whitespace normalization
- exact-content deduplication
- controlled abbreviation expansion
Correct Answer: C
Correct Answer
Answer C is correct because exact-content deduplication is appropriate because it will remove repeated identical text blocks before assembling the model context. The requirement centers on a situation where duplicated passages waste tokens and can overweight one source.
Incorrect Answers
Answer A is incorrect because date-format canonicalization can help a pipeline convert multiple source date conventions into one explicit machine-readable format before model consumption, especially when ambiguous date strings can be interpreted differently across locales. The present requirement is duplicated passages waste tokens and can overweight one source, so it is not the best answer.
Answer B is incorrect because whitespace normalization would improve another stage by enabling it to collapse accidental repeated spaces and normalize line endings while preserving semantic paragraph boundaries. It fits when formatting noise creates unnecessary tokens without carrying meaning. The stated problem is duplicated passages waste tokens and can overweight one source.
Answer D is incorrect because controlled abbreviation expansion targets a neighboring concern. It can expand organization-approved abbreviations using a deterministic dictionary before inference when domain abbreviations are ambiguous to a general model but have authoritative internal meanings. The key requirement here is duplicated passages waste tokens and can overweight one source.
Question 18
A cloud engineering team must correct a data-processing weakness. In a medical-device corpus, ‘CAPA’ must always be expanded to the company’s approved phrase before the FM summarizes incidents. Which implementation best satisfies the requirement? Choose ONE.
- navigation-boilerplate removal
- language detection and language-specific routing
- low-quality record filtering
- controlled abbreviation expansion
Correct Answer: D
Correct Answer
Answer D is correct because Select controlled abbreviation expansion. Its purpose is to expand organization-approved abbreviations using a deterministic dictionary before inference. The operational requirement is that domain abbreviations are ambiguous to a general model but have authoritative internal meanings.
Incorrect Answers
Answer A is incorrect because Using navigation-boilerplate removal would let the team strip repeated menus, cookie notices, and site chrome while retaining article content. That is appropriate when web-page boilerplate consumes context and distracts from the source information. The scenario instead calls for domain abbreviations are ambiguous to a general model but have authoritative internal meanings.
Answer B is incorrect because language detection and language-specific routing has value because it can detect the dominant language and route the record to a language-appropriate prompt or supported model path. Its proper fit is when documents arrive in several languages and the pipeline must select processing that matches each language. This question requires domain abbreviations are ambiguous to a general model but have authoritative internal meanings.
Answer C is incorrect because low-quality record filtering can help a pipeline exclude records that fail a measured quality threshold instead of allowing noisy inputs to degrade response consistency, especially when poor source quality is known and can be scored before inference. The present requirement is domain abbreviations are ambiguous to a general model but have authoritative internal meanings, so it is not the best answer.
Question 19
A regulated workflow requires deterministic preprocessing before inference. OCR output with confidence below an approved threshold produces nonsense tokens that reduce the quality of generated summaries. Which implementation best satisfies the requirement? Choose ONE.
- low-quality record filtering
- unit normalization with original-value retention
- whitespace normalization
- Unicode normalization
Correct Answer: A
Correct Answer
Answer A is correct because low-quality record filtering is the direct remedy. It can exclude records that fail a measured quality threshold instead of allowing noisy inputs to degrade response consistency. The critical scenario condition is that poor source quality is known and can be scored before inference.
Incorrect Answers
Answer B is incorrect because unit normalization with original-value retention addresses another concern by helping to convert measurements to a common unit while preserving the source value for traceability. It fits where mixed units make comparison unreliable unless values are standardized. Here the deciding need is poor source quality is known and can be scored before inference.
Answer C is incorrect because Using whitespace normalization would let the team collapse accidental repeated spaces and normalize line endings while preserving semantic paragraph boundaries. That is appropriate when formatting noise creates unnecessary tokens without carrying meaning. The scenario instead calls for poor source quality is known and can be scored before inference.
Answer D is incorrect because Unicode normalization has value because it can convert canonically equivalent Unicode sequences to one representation before deduplication and inference. Its proper fit is when visually identical text is represented by different code-point sequences. This question requires poor source quality is known and can be scored before inference.
Question 20
A developer is hardening an FM-consumption workflow. A call transcript must keep CUSTOMER and AGENT turns intact while removing nonsemantic artifacts such as duplicated ‘[noise]’ markers. Which implementation best satisfies the requirement? Choose ONE.
- speaker-turn cleanup and labeling
- Amazon Comprehend entity extraction
- unit normalization with original-value retention
- date-format canonicalization
Correct Answer: A
Correct Answer
Answer A is correct because speaker-turn cleanup and labeling is the best fit because it can remove transcription artifacts and preserve who said each utterance before passing a conversation to the FM. The workflow needs a case where a transcript contains repeated filler markers and ambiguous speaker boundaries.
Incorrect Answers
Answer B is incorrect because Amazon Comprehend entity extraction can identify entities such as organizations, people, and locations before enriching the model input. It fits when explicit entity metadata will improve downstream context and routing. This scenario instead requires a transcript contains repeated filler markers and ambiguous speaker boundaries, so it targets a different problem.
Answer C is incorrect because unit normalization with original-value retention is useful when mixed units make comparison unreliable unless values are standardized because it can convert measurements to a common unit while preserving the source value for traceability. The current requirement is a transcript contains repeated filler markers and ambiguous speaker boundaries, which makes this a mismatch.
Answer D is incorrect because date-format canonicalization addresses another concern by helping to convert multiple source date conventions into one explicit machine-readable format before model consumption. It fits where ambiguous date strings can be interpreted differently across locales. Here the deciding need is a transcript contains repeated filler markers and ambiguous speaker boundaries.
