AI-102 Implement knowledge mining and document intelligence solutions — Timed 15 Questions

Question 1

A company uses Azure Document Intelligence to extract data from invoices. They deploy the model to a container for on-premises processing. After deployment, they notice that the container consumes more memory than expected. What should they do to optimize memory usage?

15 minute time limit — choose an answer to begin.

Accepted Answer

Set the 'Memory' environment variable to a lower value in the container configuration. Azure Document Intelligence containers expose a 'Memory' environment variable that allows you to limit the container's memory allocation. By setting this variable to a lower value, you constrain the container's memory usage, which directly addresses the issue of higher-than-expected consumption. This is the recommended approach for optimizing memory in containerized deployments.

Answer

Use the 'Read' model instead of the 'Layout' model

Answer

Use the cloud API instead of the container

Answer

Reduce the batch size in the client application

Question 2

A company uses Azure Document Intelligence to extract data from tax forms. They need to improve accuracy for a specific field. Which TWO actions should they take?

Accepted Answer

Label more examples of the specific field in the training set. Labeling more examples of the specific field in the training set directly provides the custom model with additional ground-truth annotations for that field. This increases the model's ability to learn the variations in handwriting, formatting, and layout for that field, which is the most effective way to improve extraction accuracy for a targeted field in Azure Document Intelligence custom models.

Answer

Increase the batch size in the analysis request

Answer

Reduce the image resolution to 200 DPI

Answer

Use the prebuilt-tax.us model

Question 3

A company uses this skillset in an Azure AI Search enrichment pipeline. They notice that the enrichment pipeline fails when processing a document larger than 5000 characters. What is the most likely cause?

Accepted Answer

The output field mapping is missing or incorrect. The enrichment pipeline fails because the output field mapping is missing or incorrect. When a skillset processes documents, the output of each skill must be explicitly mapped to an index field; if this mapping is absent or misconfigured, the pipeline cannot store the enriched data and fails, especially for larger documents that produce more output data.

Answer

The maximum page length is too small

Answer

The default language code is not supported

Answer

The text split mode should be 'sentences'

Question 4

A company uses Azure Cognitive Search to index customer support emails. They need to implement a custom skill that extracts the sentiment of the email body and also identifies the primary product mentioned. The custom skill is a Python function deployed as an Azure Function. They want to ensure the skill can process multiple documents concurrently and handle errors gracefully. Which THREE configurations should they apply?

Accepted Answer

Set the inputs of the skill to include the 'text' field from the enriched document and the 'product' field from a previous skill.. Option B is correct because a custom skill's inputs must map the enriched document fields the Python function needs — here the 'text' field for sentiment analysis and the 'product' field produced by a previous skill — via the inputs array with name/source mappings. Option C is correct because a custom skill is invoked as an HTTP endpoint, so the uri must point to the Azure Function's URL and include the function key (e.g., ?code=...) for authentication, since Cognitive Search calls the function over HTTP. Option E is correct because the batchSize property on the custom skill controls how many documents are sent to the function in a single call, enabling concurrent/parallel processing of multiple documents. Option A is wrong because continueOnError should be set to true (not false) to handle errors gracefully and let the indexer continue when a single document fails. Option D is wrong because '/document/pages/*' context is used for splitting documents into pages (as with the Document Extraction skill), not for a custom skill that processes whole email bodies, and it would not achieve the required concurrency or error handling.

Answer

Set continueOnError to false to ensure that if the skill fails for one document, the entire indexer run stops.

Answer

Set the context property to '/document/pages/*' to process each page of the email individually.

Question 5

Refer to the exhibit. You have a skillset with two skills. You run the indexer and find that the output field 'organizations' is empty for documents that clearly contain organization names. The 'keyPhrases' output is populated correctly. What is the most likely cause of the issue?

Accepted Answer

The skill is not configured with a 'defaultLanguageCode' and the documents are not in English.. The 'keyPhrases' output is populated correctly, indicating the text extraction and overall pipeline are functional. The Entity Recognition skill requires a 'defaultLanguageCode' to correctly identify entities; if it is not set and the documents are not in English, the skill may fail to extract organizations, resulting in an empty 'organizations' field. This is a known behavior where the skill defaults to English and cannot process other languages without explicit configuration.

Answer

The skill's 'name' property is set to '#1', which is invalid.

Answer

The 'categories' property is misspelled; it should be 'entityCategories'.

Answer

The input source '/document/content' is incorrect; it should be '/document/text'.

Question 6

You are a developer for a legal firm. The firm uses Azure Cognitive Search to index legal documents. They have a custom skill that performs OCR on scanned PDFs using Azure Form Recognizer. The skill is implemented as an Azure Function. Recently, the indexer has been failing with the error: "The request was canceled due to the configured HttpClient.Timeout of 100 seconds elapsing." The documents are large (up to 200 pages each). The skill calls the Form Recognizer API asynchronously. You need to resolve the timeout issue without losing the ability to process large documents. Current configuration: batchSize = 1, maxPageSize = 100, timeout = 100 seconds. You cannot change the execution time of the Form Recognizer API. What should you do?

Accepted Answer

Decrease the batchSize to 1 and increase the timeout property in the skillset skill definition to PT300S.. The error indicates the HttpClient timeout (100 seconds) is too short for large documents processed by the custom skill. Increasing the timeout to PT300S (300 seconds) in the skill definition allows the Azure Function to wait longer for the asynchronous Form Recognizer API to complete, while keeping batchSize=1 ensures each invocation handles one document at a time, preventing overload and maintaining reliability for large PDFs.

Answer

Increase the batchSize to 10 to process more documents per invocation, reducing the number of total invocations.

Answer

Increase the maxPageSize to 500 to reduce the number of API calls to Form Recognizer.

Answer

Remove the custom skill and use the built-in OCR skill in the skillset.

Question 7

A logistics company uses Azure AI Search with an indexer and skillset to enrich shipping manifests. The skillset includes a custom skill hosted on an Azure Function that enriches each document. During a full reindex, the indexer reports that the custom skill returns HTTP 429 responses for many documents, and enriched fields are missing. You need to make the pipeline resilient without changing the Function's logic. What should you do?

Accepted Answer

Reduce the indexer batch size and configure the custom skill to handle throttling with retry and backoff.. The custom skill is being throttled, so the pipeline must reduce request pressure and retry transient failures. Decreasing the indexer batch size lowers concurrent invocations, and adding retry with backoff inside the custom skill handles HTTP 429 responses gracefully. Simply raising failure thresholds or adjusting batch sizes without retry does not recover the missing enrichment outputs.

Answer

Configure the custom skill with a batchSize and set the indexer's maxFailedItems and maxFailedItemsPerBatch to allow retries.

Answer

Add a custom skill definition with a batchSize and increase the indexer's maxFailedItemsPerBatch.

Answer

Increase the indexer batch size and set maxFailedItems to a higher value.

Question 8

You are building a knowledge mining solution with Azure AI Search. Documents are stored in Azure Blob Storage and include scanned PDFs with embedded images and text. You need to extract both the text and the textual content of images during indexing. You create a data source, an index, and an indexer. You must configure the indexer to use the built-in OCR skill. What should you do first?

Accepted Answer

Attach a cognitive services account to the skillset and enable image extraction in the indexer configuration.. The built-in OCR skill in Azure AI Search requires normalized images as input. To generate these, the indexer must be configured with imageAction set to generateNormalizedImages (or a similar option) and the skillset must be attached to a cognitive services account for billing. This allows the OCR skill to extract text from images during indexing.

Answer

Configure the indexer to use the delimitedText parsing mode and set the first line as headers.

Answer

Add a custom skill that calls the Azure AI Vision Read API directly and set the batch size to 1.

Answer

Create a separate indexer for images and use a file projection to store the extracted text.

Question 9

You are building a knowledge mining solution in Azure AI Search. Documents are ingested from Azure Blob Storage and must be enriched with a custom skill that calls an Azure Function. The Function requires an API key that must not be stored in the skillset definition or source code. The indexer runs on a schedule. You need to configure the custom skill so that the API key is supplied securely at runtime. What should you do?

Accepted Answer

Create an Azure Key Vault managed HSM and assign the Search service managed identity the Key Vault Secrets User role.. The secure approach is to let the indexer authenticate to the Azure Function by using a managed identity and to retrieve the required key from Key Vault at runtime. This avoids embedding secrets in the skillset or indexer definition and satisfies the rotation and audit requirements of a production knowledge mining pipeline.

Answer

Define the API key in the indexer's configuration object and reference it from the custom skill by using a skillset variable.

Answer

Add the API key as a base64-encoded value in the custom skill's batch size property and decode it in the Azure Function.

Answer

Store the API key in Azure Key Vault and reference the secret by using a key vault secret URI in the custom skill's HTTP headers.

Question 10

A financial services firm uses an Azure AI Search indexer with a custom WebApiSkill that calls an internal scoring API. The API occasionally returns HTTP 500 responses due to transient load. The indexer currently fails the entire run whenever the custom skill returns an error. You need to make the pipeline more resilient to these transient failures. What should you do?

Accepted Answer

Configure the custom skill's errorHandling property to ignore errors.. The WebApiSkill includes an errorHandling property that determines whether an error returned by the external endpoint fails the indexer or is recorded and ignored. Setting it to ignore allows the run to complete despite intermittent failures from the scoring API. Batch size, failed item thresholds, and caching affect performance or tolerance but not how skill-level errors are treated.

Answer

Enable the indexer's cache and rerun the indexer.

Answer

Set the batchSize property of the indexer to 1.

Answer

Increase the indexer maxFailedItems value to a large number.

Question 11

A research organization uses Azure AI Search with a skillset to enrich scientific papers. They want to extract named entities such as people, organizations, and locations from the paper text and store them in a collection field in the index. The entities must be searchable individually. Which configuration should they use in the index definition?

Accepted Answer

A Collection(Edm.String) field with filterable and facetable attributes enabled.. Entity extraction yields multiple values per document, so the index needs a collection field. A Collection(Edm.String) with filterable and facetable attributes stores each entity as a separate searchable value and supports filtering and faceting by entity. Concatenated strings, complex types, and retrievable-only fields do not satisfy individual searchability and filtering needs.

Answer

A Collection(Edm.ComplexType) field with one subfield per entity type.

Answer

A single Edm.String field marked as retrievable only.

Answer

A single Edm.String field with the entities concatenated into one value.

Question 12

You are implementing a knowledge mining solution using Azure AI Search. You have an indexer that processes documents from Azure Blob Storage. You need to enrich the documents with entities extracted by Azure AI Language. You create a skillset with the Entity Recognition skill. You want to ensure that the extracted entities are stored in a complex collection field in the index. What should you do?

Accepted Answer

In the index, define a field of type Collection(Edm.ComplexType) with subfields for name, type, and confidence. In the indexer, map the skill output to this field using an output field mapping.. To store the structured output of the Entity Recognition skill in Azure AI Search, the index must include a field of type Collection(Edm.ComplexType) with subfields that match the entity properties (name, type, confidence). The indexer output field mapping then maps the skill output to this complex field, enabling advanced querying on entity attributes.

Answer

Store the raw JSON output of the skill in a string field and parse it at query time using a custom analyzer.

Answer

In the index, define a field of type Collection(Edm.String) and concatenate the entity names into a single string.

Answer

Use a shaper skill to flatten the entity collection into individual fields for each entity type, then map each to a separate string field.

Question 13

You are building a knowledge mining solution with Azure AI Search. The data source is an Azure Blob Storage container holding scanned PDF invoices. You configure a skillset that includes the built-in OCR skill followed by the built-in Entity Recognition skill. After running the indexer, you notice that entities are extracted from the OCR text but the original document content is not searchable. You need to ensure that both the raw text and the extracted entities are searchable. What should you do?

Accepted Answer

Add an output field mapping in the indexer that maps the OCR skill output to an index field of type Edm.String.. When a skillset enriches documents, the outputs exist only in the enrichment tree unless you map them to index fields. The OCR skill emits text that must be explicitly mapped. Adding an output field mapping from the OCR skill output to an Edm.String index field makes that text searchable, while the existing entity mappings continue to work. Other options do not persist the OCR text to the index.

Answer

Increase the indexer's batch size to ensure all documents are processed in a single run.

Answer

Enable the indexer's 'indexing parameters' to include the 'content' field from the data source.

Answer

Change the Entity Recognition skill to use a different language model that also outputs the source text.

Question 14

A financial services firm uses Azure AI Search to index earnings call transcripts stored in Azure Cosmos DB. They need to enrich each transcript with a custom skill that calls an external NLP API. The API returns a large JSON payload that must be split into separate fields for sentiment, entities, and topics. The indexer runs daily. You need to configure the skillset to map the API response into the index. What should you do?

Accepted Answer

Add a custom skill that parses the JSON response and outputs separate string fields, then map each field to the index in the skillset's output field mappings.. To split a complex JSON response into separate index fields, the enrichment pipeline must produce distinct output nodes. A custom skill that parses the JSON and returns separate outputs creates those nodes, and output field mappings then project each node to a corresponding index field. This keeps the index schema aligned with the NLP API's response structure.

Answer

Configure the custom skill's output with a targetName for each JSON property and map each target to an index field in the output field mappings.

Answer

Use a conditional skill to evaluate the JSON response and route each property to a separate output field.

Answer

Add a ShaperSkill to the skillset and set its inputs to the custom skill's output, then map the shaped output to the index fields.

Question 15

A financial services company uses Azure AI Document Intelligence with a custom neural model to extract fields from loan applications. The model was trained on 200 labeled documents and performs well on the training set but poorly on new documents from a different branch office. The new documents use slightly different terminology and layout. You need to improve the model's accuracy on the new documents with minimal labeling effort. What should you do?

Accepted Answer

Use the model's compose capability to combine the existing model with a new model trained on a small set of new branch documents.. Composing models allows a single endpoint to route documents to the most appropriate custom model. By training a new model on a small labeled sample from the new branch and composing it with the original, you extend coverage to the new terminology and layout while preserving the original model's accuracy. This approach requires far fewer labels than retraining from scratch.

Answer

Increase the model's confidence threshold to filter out low-confidence extractions from the new documents.

Answer

Add the new branch documents to the training set and retrain the existing model without any new labels.

Answer

Retrain the model from scratch using only the new branch office documents.