AI-102 Designing and Implementing a Microsoft Azure AI Solution Questions and Answers

Azure Metrics Advisor builds on Anomaly Detector but adds root cause analysis, correlation across multiple signals, incident detection, and alerting out-of-the-box.

D. Anomaly Detector → Works for single-signal anomaly detection but lacks correlation and root cause analysis.

C. Azure Machine Learning → Would work but requires custom model development (more dev effort).

B. Form Recognizer → Used for document processing, not sensor telemetry.

Correct Answer: A

Comprehensive Detailed ExplanationIn the provided LUDown file snippet:

Intents represent the purpose or action the user wants to perform.

Examples: ## Confirm, ## ExtractName, ## Logout, ## SelectItem, ## SelectNone.

These are all intent definitions.

So, SelectItem is an intent.

Utterances are the example phrases that users might say to invoke a particular intent.

Under ## SelectItem, one example utterance is:

choose {@DirectionalReference=top right}

This is an example phrase that maps to the SelectItem intent and contains the DirectionalReference entity.

Therefore, Choose {@DirectionalReference=top right} is an utterance.

Domain: Predefined category grouping multiple intents (not applicable here).

Entity: Extracted data within utterances (e.g., @DirectionalReference=top right or @personName).

The entire phrase "choose {@DirectionalReference=top right}" is not an entity but an utterance containing an entity.

Why not the other options?

SelectItem → an intent

Choose {@DirectionalReference=top right} → an utterance

Correct Answers

LUDown file format for LUIS

LUIS concepts: intents, utterances, and entities

Microsoft References

Analysis:

Azure AI Bot Service (A): Helps build chatbots but does not store grounding data → ❌

Azure SQL (B): Possible storage, but not recommended for vector search grounding → ❌

Azure AI Document Intelligence (C): Extracts structured info, not used directly for grounding → ❌

Azure Blob Storage (D): Required for storing raw files used in retrieval-augmented generation (RAG) → ✅

Azure AI Search (E): Used to index, vectorize, and query data for grounding → ✅

Correct Answer:

D. Azure Blob Storage and E. Azure AI Search

curl -i -X: The POST method is appropriate because you are creating a new custom analyzer, which involves sending a request to create a resource. POST is used for creating new resources in REST APIs.

https:// .services.ai.azure.com/: The base URL should be updated to include the contentunderstanding endpoint, as this is the specific service within the Azure AI Content Understanding resource (cu1) that handles analyzer-related operations. Replace with the actual resource name or endpoint provided by your Azure subscription.

/analyzers/myInvoice7api-version-2024-12-01-preview: This part of the URL is correct as it specifies the endpoint for creating a custom analyzer (/analyzers) with the name myInvoice7 and the API version 2024-12-01-preview. No changes are needed here, assuming the analyzer name and version are intended as shown.

Explanation

Query successful

An analysis of the provided code segment is needed to determine the truth value of the statements, which are not explicitly provided in the prompt. Since I don't have the statements, I will analyze the code and infer common types of questions related to this code structure to provide a helpful response.

Code AnalysisThe code segment is written in a C#-like syntax and iterates through a collection named brands.

C#

foreach (var brand in brands)

{

if (brand.Confidence >= .75)

Console.WriteLine($"Logo of {brand.Name} between {brand.Rectangle.X}, {brand.Rectangle.Y} and {brand.Rectangle.W}, {brand.Rectangle.H}");

}

Iteration: The foreach (var brand in brands) loop processes each item (presumably a detected brand or logo) in the brands collection.

Filtering/Condition: The if (brand.Confidence >= .75) statement acts as a filter. It only processes a brand if its associated Confidence value (likely a floating-point number between 0.0 and 1.0) is greater than or equal to 0.75.

Output: For brands that pass the confidence threshold, a line of text is printed to the console using string interpolation ($"").

It reports the Name of the brand.

It reports the bounding box coordinates, presumably X (horizontal position), Y (vertical position), W (width), and H (height), which are properties of the nested brand.Rectangle object.

Comprehensive Detailed ExplanationYou trained an object detection model obj1 inside project proj1 on the acvdev Custom Vision resource and need to move it to the acvprod resource.

Moving a model between Custom Vision resources is done by exporting the project (not an iteration) from the source resource and importing it into the target resource.

Step 1: GetProjects (acvdev)Retrieve the list of projects on the development resource to obtain the project ID of proj1. The export operation requires the project ID.

Step 2: ExportProject (acvdev)Call the Export Project API on the development resource to generate the portable project package (includes tags, images’ metadata, iterations, settings). This is specifically intended for moving/cloning a project across resources or regions.

Step 3: ImportProject (acvprod)Use the Import Project API on the production resource to create a new project from the exported package. This recreates the project (and associated iterations/metadata) under acvprod.

Notes:

ExportIteration is for exporting a trained iteration to formats for edge/container or mobile use, not for transferring to another Custom Vision resource.

UpdateProject is for changing metadata (name, description, domain) of an existing project and is not part of the move workflow.

Custom Vision REST API – Projects: Export/Importhttps://learn.microsoft.com/azure/ai-services/custom-vision-service/export-import-project

Custom Vision REST API – Projects API (Get projects)https://learn.microsoft.com/rest/api/customvision/training/projects

Custom Vision – Export model vs. Export project guidancehttps://learn.microsoft.com/azure/ai-services/custom-vision-service/export-your-model (describes iteration export purpose versus project portability)

References (Microsoft Azure AI Solution)

Active learning in Language Understanding (LUIS / Conversational Language Understanding) works by collecting real user queries sent to your prediction endpoint.

By setting log=true in the prediction endpoint query, these user utterances are stored in the LUIS service.

The logged utterances can then be reviewed in the Review endpoint utterances page of the LUIS portal, where you can relabel and retrain the model. This enables the active learning loop.

Comprehensive Detailed ExplanationWhy not the others?

A. show-all-intents=true: This simply forces the endpoint to return all intents, not related to active learning.

B. Enable speech priming: Used for improving speech-to-text accuracy, unrelated to LUIS active learning.

D. Enable sentiment analysis: Adds sentiment scores to prediction output, not related to active learning.

You are building a chatbot that will use an Azure OpenAI model.

You need to deploy the model in your Azure subscription.

Apply for access to Azure OpenAI

Azure OpenAI is a gated service and requires approval before you can provision resources.

Provision an Azure OpenAI resource

After approval, you must provision an Azure OpenAI resource in your subscription.

Deploy the GPT model

Since the chatbot requires natural language understanding and conversation, you deploy a GPT model (e.g., GPT-4, GPT-3.5).

Key Steps to Deploy an Azure OpenAI Model for a Chatbot:

Provision Azure API Management → Not needed for basic deployment (used only for advanced API management scenarios).

Deploy the embeddings model → Used for search/RAG, not necessary for chatbot baseline.

Deploy the DALL-E model → Used for image generation, not chatbot text responses.

Actions That Are Not Required for this Question:

Apply for access to Azure OpenAI.

Provision an Azure OpenAI resource.

Deploy the GPT model.

Correct Order:

Verified Answer (Sequence):

Apply for access to Azure OpenAI

Provision an Azure OpenAI resource

Deploy the GPT model

Microsoft References:

Quickstart: Get started with Azure OpenAI

Deploying models in Azure OpenAI

Comprehensive Detailed ExplanationThe requirement is to build an Adaptive Card for the chatbot that supports multilingual product display and meets accessibility requirements.

Product Name Display

The product JSON stores localized names like "name": { "en": "...", "es": "...", "pt": "..." }.

To dynamically render based on the current language, you must use name[language].

This ensures the product name shown in the card adapts automatically to the user’s preferred language.

Stock Level Warning

Business rules require warnings when stock is low or out of stock.

This means you display the stock warning only when stockLevel is not 'OK'.

Therefore, the correct condition is:"$when": "${stockLevel != 'OK'}"

Image Alt Text for Accessibility

Accessibility requirements specify that all images must have alt text in English, Spanish, and Portuguese.

In the product JSON, "alttext" is also localized:

"altText": { "en": "Bicycle", "es": "Bicicleta", "pt": "Bicicleta" }

To render the correct localized alt text dynamically, use image.altText[language].

Correct Selections:

First blank: name[language]

Second blank: "$when": "${stockLevel != 'OK'}"

Third blank: image.altText[language]

Adaptive Cards Templating

Adaptive Card $when property

Multilingual data in Azure Cognitive Search and Adaptive Cards

Microsoft References

Parameter type: stream

Visual features list: VisualFeatureTypes.Description

Result to use: results.Description.Captions[0]

To generate accessible alt text, you should use the caption produced by Azure Computer Vision’s Description feature (it produces a human-readable sentence with a confidence score). Therefore:

The image input should be a stream, because you’re uploading images (not just passing URLs) during product creation and AnalyzeImage…Async supports image streams.

Request the Description feature in VisualFeatureTypes so the service returns results.Description.Captions.

Use results.Description.Captions[0] and return the caption text if its confidence is high enough (e.g., > 0.5) to meet the accessibility requirement that all images must have relevant alt text.

Other features (Tags, Objects, Brands) are useful for enrichment but do not directly return natural-language captions suitable for alt text.

Microsoft Azure AI Solution References

Computer Vision (Image Analysis) – Description/Captions and features: Microsoft Docs, Image Analysis – VisualFeatureTypes.Description returns description.captions.https://learn.microsoft.com/azure/ai-services/computer-vision/concept-image-analysis

SDK usage (analyze image from a stream): Microsoft Docs, Analyze an image by using the Computer Vision client library.https://learn.microsoft.com/azure/ai-services/computer-visi on/how-to/call-analyze-image?tabs=version-3-2#analyze-an-image-from-a-stream

When using Azure Translator REST API, required headers are:

Ocp-Apim-Subscription-Key → subscription key

Ocp-Apim-Subscription-Region → resource region (if using a multi-service Cognitive Services resource)

Content-Type → typically application/json or application/xml

Other options:

Client trace ID is optional for debugging.

Resource ID, access control request, content language, content length are not required headers.

Correct Answer: B

Non-relational databases (NoSQL) typically store semi-structured or unstructured data and have the following characteristics:

Entities are often self-describing, meaning they store their own schema within the data (e.g., JSON documents).

Schema is flexible, allowing for changes without restructuring the entire database.

Other options:

Full support for Transact-SQL → This is a characteristic of relational (SQL) databases, not NoSQL.

A fixed schema → Also a relational database characteristic.

Correct Answer: self describing entities

Within a search service, synonym maps are a global resource that associate equivalent terms, expanding the scope of a query without the user having to actually provide the term. For example, assuming "dog", "canine", and "puppy" are mapped synonyms, a query on "canine" will match on a document containing "dog".

Create synonyms: A synonym map is an asset that can be created once and used by many indexes.

Box 1: api-nam.cognitive.microsofttranslator.com

https://docs.microsoft.com/en-us/azure/cognitive-services/translator/referenc e/v3-0-reference

Box 2: /translate

Scenario: Support autocompletion and autosuggestion based on all product name variants.

A: Call a suggester-enabled query, in the form of a Suggestion request or Autocomplete request, using an API. API usage is illustrated in the following call to the Autocomplete REST API.

POST /indexes/myxboxgames/docs/autocomplete?search&api-version=2020-06-30

{

"search": "minecraf",

"suggesterName": "sg"

}

B: In Azure Cognitive Search, typeahead or "search-as-you-type" is enabled through a suggester. A suggester provides a list of fields that undergo additional tokenization, generating prefix sequences to support matches on partial terms. For example, a suggester that includes a City field with a value for "Seattle" will have prefix combinations of "sea", "seat", "seatt", and "seattl" to support typeahead.

F. Use the default standard Lucene analyzer ("analyzer": null) or a language analyzer (for example, "analyzer": "en.Microsoft") on the field.

This question is about selecting valid training files for a custom Form Recognizer model in Azure AI Document Intelligence (formerly Form Recognizer).

Supported File Formats for TrainingAccording to Microsoft documentation, Form Recognizer supports the following for training and analysis:

PDF (≤ 500 pages, ≤ 50 MB for training)

JPEG/JPG (≤ 50 MB)

PNG (≤ 50 MB)

TIFF (≤ 50 MB)

Unsupported:

MP4 → Video not supported.

GIF → Not supported.

Images (JPG, PNG, TIFF): ≤ 50 MB

PDFs: ≤ 50 MB

Note: 500 pages limit applies to PDFs.

Maximum File Size Rules (for training input files):

File1 – PDF – 20 MB

Type supported.

Size within limit (< 50 MB).

Valid ✅

File2 – MP4 – 100 MB

MP4 is not supported.

Invalid ❌

File3 – JPG – 20 MB

Type supported.

Size within limit (< 50 MB).

Valid ✅

File4 – PDF – 100 MB

Type supported.

Size exceeds the 50 MB maximum limit.

Invalid ❌

File5 – GIF – 1 MB

GIF format not supported.

Invalid ❌

File6 – JPG – 40 MB

Type supported.

Size within limit (< 50 MB).

Valid ✅

File-by-File Analysis

Correct Answer:

A. File1, C. File3, F. File6

Azure AI Document Intelligence (Form Recognizer) – Supported file formats

Train a custom model with Azure Form Recognizer

Microsoft References

Comprehensive Detailed ExplanationThe question asks about testing a bot interactively on a local machine when using the Microsoft Bot Framework SDK.

Steps to test locally:

Build and run the bot

Compile the bot application code and run it locally (e.g., using Visual Studio or Node.js).

This exposes the bot’s endpoint on localhost (commonly http://localhost:3978/api/messages ).

Open the Bot Framework Emulator

The Bot Framework Emulator is a desktop app for testing and debugging bots built with the Bot Framework SDK.

It allows you to send messages to the bot and see the responses.

Connect to the bot endpoint

Use the Emulator to connect to the locally running bot by entering its endpoint URL.

Once connected, you can interact with the bot as if you were a real user.

Register the bot with the Azure Bot Service → Needed for cloud deployment, not local testing.

Open the Bot Framework Composer → Only needed if designing the bot with Composer, but this scenario uses the SDK directly.

Actions Not Required in This Case

Build and run the bot → Open the Bot Framework Emulator → Connect to the bot endpoint

Correct Sequences (multiple possible valid orders):

Test a bot locally with Bot Framework Emulator

Bot Framework SDK overview

Microsoft References

✅ Final Answer Sequence:

Build and run the bot

Open the Bot Framework Emulator

Connect to the bot endpoint

Comprehensive Detailed ExplanationYou are tasked with building a solution using Custom Vision to detect faults in factory components. Custom Vision follows a defined workflow:

Create a project

Before anything else, you must create a Custom Vision project in the Azure AI portal.

This project specifies the type of model you are building: classification (classifying entire images) or object detection (locating items within images).

Since the question is about detecting faults in specific components (a business-specific scenario), it aligns with classification unless explicitly stated otherwise.

Upload and tag images

You must provide training data.

Upload a set of images showing both faulty and non-faulty components.

Tag each image correctly so that the model can learn from labeled examples.

Train the classifier model

Once images are tagged, you run training on the dataset to generate the classifier model.

The model can then be published and used for inference via API calls.

Initialize the training dataset: This is not a direct step in Custom Vision; instead, uploading and tagging images inherently prepares the training dataset.

Train the object detection model: Object detection is used if you need bounding boxes around items within an image. Since the question is about recognizing faults in a component (faulty vs non-faulty), a classifier model is appropriate.

Why not the other options?

Correct Sequence:

Create a project

Upload and tag images

Train the classifier model

Quickstart: Build an image classification project with the Custom Vision portal

Custom Vision overview

Train and test a Custom Vision model

Microsoft References

When working with Azure OpenAI Studio to configure a chatbot or conversational AI model, two key parameters control the creativity vs determinism of responses:

Temperature

Controls the amount of randomness in the model’s output.

A lower value (closer to 0) makes the output more deterministic and focused.

A higher value increases creativity and randomness.

For deterministic responses, you should reduce this value (for example, 0.2 or 0.0).

Top P (nucleus sampling)

Controls the diversity of responses by limiting the probability space considered.

A lower value (closer to 0) makes the responses more deterministic by narrowing the token choices.

A higher value increases creativity by expanding possible outputs.

For deterministic responses, keep this value low (e.g., 0.1–0.3).

Max response: Sets maximum token length of the output, not determinism.

Stop sequence: Defines custom stopping conditions, not randomness.

Frequency penalty and Presence penalty: Control repetition and encouragement of new topics, not determinism.

Why Not Other Parameters?Thus, the correct parameters to adjust for more deterministic and less creative responses are:

Temperature and Top P.

Correct Answer: Temperature, Top P

Azure OpenAI Service – Parameters

Azure OpenAI Studio Overview

Nucleus Sampling (Top P) explanation

Microsoft References

API Endpoint SelectionThe constraint states that the content of the text must remain within the Americas Azure geography.

"api.cognitive.microsofttranslator.com": This is the global endpoint, which doesn't guarantee data residency in a specific geography.

"api-apc.cognitive.microsofttranslator.com": This endpoint refers to the Asia Pacific geography.

"api-nam.cognitive.microsofttranslator.com": This endpoint refers to the North and South America (NAM) geography.

Therefore, to meet the data residency requirement for the Americas, you must select "api-nam.cognitive.microsofttranslator.com".

2. Path/Query String SelectionThe requirement is to translate the text to a single language. The input text can be in "one of many languages," meaning the source language isn't fixed and should be detected automatically. The language to translate to is implied to be English (en) based on the options available and common translation scenarios.

"/translate?from=en": This is incorrect because it specifies the source language as English, which contradicts the "text can be in one of many languages" requirement (implying automatic source language detection).

"/translate?suggestedFrom=en": This is not the standard way to initiate a translation request and seems to suggest a source language, which is unnecessary if auto-detection is desired/possible.

"/translate?to=en": This is the correct and standard way to request a translation where the source language is automatically detected, and the target language is English (en). This fulfills the translation requirement.

"/detect?to=en" / "/detect?from=en": These endpoints are used for language detection and not for the translation itself. The prompt explicitly asks to develop code to translate the text.

When testing a bot with the Bot Framework Emulator, especially if the bot uses authentication (OAuth), you need the Emulator to mimic real-world authentication flows. To do that:

Use version 1.0 authentication tokens

Ensures that authentication tokens are generated in the same format as Bot Framework v1.0, which is compatible with most OAuth flows.

Use a sign-in verification code for OAuthCard

Prompts the user with a verification code during sign-in flows, ensuring the emulator behaves like a real client app (e.g., mobile or Teams).

This is required to test sign-in prompts and enforce credential entry.

Use your own user ID to communicate with the bot

Ensures that the Emulator identifies you uniquely and triggers authentication properly.

Without this, the bot may treat sessions generically, and authentication prompts might not behave as expected.

Why not the others?

Application Updates (automatically download/install) → Related to emulator updates, not authentication.

Data Collection → Only for telemetry; does not affect authentication.

Ngrok settings → Used for tunneling remote bot connections; unrelated to credential prompts.

Correct Answer Selections:

Use version 1.0 authentication tokens

Use a sign-in verification code for OAuthCard

Use your own user ID to communicate with the bot

Bot Framework Emulator - Authentication

Testing bots with Bot Framework Emulator

Microsoft References

The system message you provided—“You are an AI assistant that helps people solve mathematical puzzles. Explain your answers as if the request is by a 4-year-old.”—is classic priming (also called system prompt/role prompting). Priming sets the assistant’s persona, objectives, tone, and constraints before any user message, steering all subsequent responses accordingly.

Why the other options are incorrect:

A. Few-shot learning – supplies example Q&A pairs in the prompt to teach the model by example; none were included here.

B. Affordance – relates to interface cues that guide users; not a prompt technique for model behavior.

C. Chain of thought – asks the model to show intermediate reasoning steps; your prompt doesn’t request step-by-step rationale, only a style/tone.

Microsoft Azure AI References

Azure OpenAI – Prompt engineering techniques (role/system prompts, priming).

Azure OpenAI – System message (role) guidance for controlling assistant behavior and tone.

The requirement: Ensure the chatbot blocks questions that attempt to circumvent built-in safety features.

This is known as a jailbreak attempt (e.g., prompt injection, role-play tricking).

Azure AI Content Safety recently added Jailbreak risk detection for precisely this purpose.

Other options:

Protected material detection → for copyright/IP detection.

Monitor online activity → not related to OpenAI safety.

Moderate text content → general harmful content moderation (violence, hate, sexual, self-harm), but not jailbreaks.

Microsoft Reference:

Azure AI Content Safety jailbreak risk detection

Scenario: All videos must have transcripts that are associated to the video and included in product descriptions.

Product descriptions, transcripts, and all text must be available in English, Spanish, and Portuguese.

Step 1: Upload the video to blob storage

Given a video or audio file, the file is first dropped into a Blob Storage. T

Step 2: Index the video by using the Video Indexer API.

When a video is indexed, Video Indexer produces the JSON content that contains details of the specified video insights. The insights include: transcripts, OCRs, faces, topics, blocks, etc.

Step 3: Extract the transcript from the Video Indexer API.

Step 4: Translate the transcript by using the Translator API.

Box 1: Azure Blob storage

At the start of the pipeline, you have unstructured text or non-text content (such as images, scanned documents, or JPEG files). Data must exist in an Azure data storage service that can be accessed by an indexer.

Box 2: Computer Vision API

Scenario: Provide users with the ability to search insight gained from the images, manuals, and videos associated with the products.

The Computer Vision Read API is Azure's latest OCR technology (learn what's new) that extracts printed text (in several languages), handwritten text (English only), digits, and currency symbols from images and multi-page PDF documents.

Box 3: Translator API

Scenario: Product descriptions, transcripts, and all text must be available in English, Spanish, and Portuguese.

Box 4: Azure Files

Scenario: Store all raw insight data that was generated, so the data can be processed later.

Box 1: QnA Maker Editor

Scenario: Provide all employees with the ability to edit Q&As.

The QnA Maker Editor (read/write) has the following permissions:

Create KB API

Update KB API

Replace KB API

Replace Alterations

"Train API" [in

new service model v5]

Box 2: Contributor

Scenario: Only senior managers must be able to publish updates.

Contributor permission: All except ability to add new members to roles

The wiki contains text in English, French and Portuguese.

Scenario: All planned projects must support English, French, and Portuguese.

The Document Extraction skill extracts content from a file within the enrichment pipeline. This allows you to take advantage of the document extraction step that normally happens before the skillset execution with files that may be generated by other skills.

Note: The Translator Text API will be used to determine the from language. The Language detection skill is not required.

Box 1: "categories": ["Locations", "Persons", "Organizations"],

Locations, Persons, Organizations are in the outputs.

Scenario: Contoso plans to develop a searchable knowledgebase of all the intellectual property

Note: The categories parameter is an array of categories that should be extracted. Possible category types: "Person", "Location", "Organization", "Quantity", "Datetime", "URL", "Email". If no category is provided, all types are returned.

Box 2: {"name": " entities"}

The include wikis, so should include entities in the outputs.

Note: entities is an array of complex types that contains rich information about the entities extracted from text, with the following fields

name (the actual entity name. This represents a "normalized" form)

wikipediaId

wikipediaLanguage

wikipediaUrl (a link to Wikipedia page for the entity)

etc.

C: Analyze Receipt - Get Analyze Receipt Result.

Query the status and retrieve the result of an Analyze Receipt operation.

Request URL: https://{endpoint}/formrecognizer/v2.0-preview/prebuilt/receipt/analyzeResults/{resultId}

E: POST {Endpoint}/vision/v3.2/read/analyze

Use this interface to get the result of a Read operation, employing the state-of-the-art Optical Character Recognition (OCR) algorithms optimized for text-heavy documents.

Scenario: Contoso plans to develop a document processing workflow to extract information automatically from PDFs and images of financial documents

The document processing solution must be able to process standardized financial documents that have the following characteristics:

- Contain fewer than 20 pages.

- Be formatted as PDF or JPEG files.

- Have a distinct standard for each office.

*The document processing solution must be able to extract tables and text from the financial documents.

The document processing solution must be able to extract information from receipt images.

Step 1: Create a new project and load sample documents

Create a new project. Projects store your configurations and settings.

Step 2: Label the sample documents

When you create or open a project, the main tag editor window opens.

Step 3: Train a custom model.

Finally, train a custom model.

Scenario: All planned projects must support English, French, and Portuguese.

If a bot uses multiple LUIS models and QnA Maker knowledge bases (knowledge bases), you can use the Dispatch tool to determine which LUIS model or QnA Maker knowledge base best matches the user input. The dispatch tool does this by creating a single LUIS app to route user input to the correct model.

The requirements state:

All AI solution responses must have a confidence score ≥ 70%.

If the response confidence score is < 70%, the response must be improved by human input.

Members of the Consultant-Bookkeeper group must be able to process financial documents, which includes performing manual review when the AI confidence is below threshold.

Looking at the provided JSON:

"fields": {

"ReceiptType": {

"type": "string",

"valueString": "Itemized",

"confidence": 0.672

},

"MerchantName": {

"type": "string",

"valueString": "Tailwind",

"confidence": 0.913

}

}

ReceiptType.confidence = 0.672 → this is below 0.7.

MerchantName.confidence = 0.913 → this is above 0.7.

Therefore, the correct condition for triggering manual review is:

documentResults.fields..confidence < 0.7

Now, evaluating the options:

A. documentResults.docType == "prebuilt:receipt"

Always true for receipts, does not check confidence. Not correct.

B. documentResults.fields.*.confidence < 0.7

This is the correct general expression: trigger manual review whenever any field confidence is below 0.7.

C. documentResults.fields.ReceiptType.confidence > 0.7

This would bypass manual review when ReceiptType has high confidence. The requirement is to trigger review when confidence < 0.7, so this is the opposite.

D. documentResults.fields.MerchantName.confidence < 0.7

This only checks one field (MerchantName). In the JSON, MerchantName has confidence 0.913 (>0.7), so this condition would not trigger, but ReceiptType clearly needs review. Too narrow.

Correct Answer: B. documentResults.fields.*.confidence < 0.7

Azure AI Document Intelligence – Confidence scores

Human-in-the-loop for Document Intelligence

Microsoft References

Box 1: Cognitive Service User

Ensure that the members of a group named Management-Accountants can approve the FAQs.

Approve=publish.

Cognitive Service User (read/write/publish): API permissions: All access to Cognitive Services resource except for ability to:

1. Add new members to roles.

2. Create new resources.

Box 2: Cognitive Services QnA Maker Editor

Ensure that the members of a group named Consultant-Accountants can create and amend the FAQs.

QnA Maker Editor: API permissions:

1. Create KB API

2. Update KB API

3. Replace KB API

4. Replace Alterations

5. "Train API" [in new service model v5]

Box 3: Cognitive Services QnA Maker Read

Ensure that the members of a group named the Agent-CustomerServices can browse the FAQs.

QnA Maker Read: API Permissions:

1. Download KB API

2. List KBs for user API

3. Get Knowledge base details

4. Download Alterations

Generate Answer

Can search content in different formats, including video

Audio and video insights (multi-channels). When indexing by one channel, partial result for those models will be available.

Keywords extraction: Extracts keywords from speech and visual text.

Named entities extraction: Extracts brands, locations, and people from speech and visual text via natural language processing (NLP).

Topic inference: Makes inference of main topics from transcripts. The 2nd-level IPTC taxonomy is included.

Artifacts: Extracts rich set of "next level of details" artifacts for each of the models.

Sentiment analysis: Identifies positive, negative, and neutral sentiments from speech and visual text.

Explanation:

Scenario: When the response confidence score is low, ensure that the chatbot can provide other response options to the customers.

When no good match is found by the ranker, the confidence score of 0.0 or "None" is returned and the default response is "No good match found in the KB". You can override this default response in the bot or application code calling the endpoint. Alternately, you can also set the override response in Azure and this changes the default for all knowledge bases deployed in a particular QnA Maker service.

Choosing Ranker type: By default, QnA Maker searches through questions and answers. If you want to search through questions only, to generate an answer, use the RankerType=QuestionOnly in the POST body of the GenerateAnswer request.