AI-102 · topic practice

Implement natural language processing solutions practice questions

Q: How should I use these Implement natural language processing solutions practice questions?

Read each scenario carefully and choose your answer before revealing the explanation. Then check why your choice was right or wrong. Repeat until the reasoning feels automatic.

Q: Can I practise just Implement natural language processing solutions questions in a focused session?

Yes — use the session launcher on this page to start a 10-, 20-, 30- or 50-question session drawn entirely from the Implement natural language processing solutions domain.

This domain covers Azure AI Language and Azure AI Speech: CLU and question answering, sentiment and key phrase extraction, translation, speech-to-text and text-to-speech. Questions are scenario-based, asking you to pick the right service, language support strategy, authentication method, or network configuration for a stated requirement.

Courseiva uses original exam-style practice questions designed for learning and revision. The goal is to understand the concepts, recognise exam patterns, and improve through explanations — not memorise copied exam dumps.

Editorial oversight:Johnson Ajibi· MSc IT Security, IEEE Senior Member

20 questionsDomain: Implement natural language processing solutions

Practice 10 questions Browse domain →

What the exam tests

What to know about Implement natural language processing solutions

Be able to select the correct Azure AI Language or Speech feature for a scenario, configure multilingual intent recognition, and secure access to the resource. The most important thing is matching the service and its language or network configuration to the stated requirement.

Choosing Azure AI Language features such as CLU, question answering, NER, and sentiment analysis for a scenario

Configuring multilingual intent recognition, including per-language models versus multilingual utterance training

Using Azure AI Speech batch transcription, real-time transcription, and custom speech models

Securing Language and Speech resources with keys, Microsoft Entra ID, managed identities, and network rules

Why learners struggle

Why Implement natural language processing solutions questions are commonly missed

NAT questions are missed when learners confuse the four address types (inside local, inside global, outside local, outside global) or misapply the interface direction. A translation rule can look correct but still fail if the ACL, interface, or direction is wrong.

·Inside local vs inside global — inside local is the private source, inside global is the translated public address
·PAT overloads — many sources share one public IP using unique port numbers
·Interface direction — ip nat inside and ip nat outside must be on the correct interfaces
·Static NAT vs dynamic NAT vs PAT — each serves a different use case
·The NAT ACL identifies traffic to translate, not traffic to permit or deny
·A missing translation can look like a routing problem if the interfaces are misconfigured

Watch out for

Common Implement natural language processing solutions exam traps

▸Assuming one LUIS or CLU model automatically handles all languages; multilingual support requires explicit configuration or separate language models.
▸Confusing batch transcription with real-time speech-to-text when latency requirements are stated.
▸Forgetting that allowing on-premises access requires configuring firewall, virtual network, or private endpoint settings, not just copying a key.

Practice set

Implement natural language processing solutions questions

20 questions · select your answer, then reveal the explanation

Question 1easymultiple choice

Read the full Implement natural language processing solutions explanation →

A company is building a chatbot using Azure Cognitive Service for Language. They need to ensure that user utterances are correctly mapped to the appropriate intent in a custom question answering project. What should they configure?

Trap 1: Add synonyms and phrase list to the LUIS application.

LUIS is not used in custom question answering; synonyms are configured directly in the knowledge base.

Trap 2: Add alternative question phrases to the QnA pairs in the custom…

While adding alternative questions is useful, the question specifically asks about synonyms and phrase lists, not alternative questions.

Trap 3: Define entities in the custom question answering project to capture…

Entities are used in LUIS and conversational language understanding, not in custom question answering.

Study all Implement natural language processing solutions common traps →

A
Add synonyms and phrase list to the LUIS application.
Why it fails: LUIS is not used in custom question answering; synonyms are configured directly in the knowledge base.
B
Add alternative question phrases to the QnA pairs in the custom question answering project.
Why it fails: While adding alternative questions is useful, the question specifically asks about synonyms and phrase lists, not alternative questions.
C
Define entities in the custom question answering project to capture key information.
Why it fails: Entities are used in LUIS and conversational language understanding, not in custom question answering.
D
Add synonyms and phrase list to the custom question answering project.
Synonyms and phrase lists help match varied user utterances to the correct QnA pair.

Full breakdown with real-world context →

Question 2mediummultiple choice

Read the full Implement natural language processing solutions explanation →

A development team is using the Azure Cognitive Service for Language to perform sentiment analysis on social media posts. They notice that the returned sentiment scores are often neutral for posts that are clearly positive or negative. What is the most likely reason?

Trap 1: The service does not support sentiment analysis for social media…

The service supports social media language.

Trap 2: The service is not configured to detect mixed sentiment.

Mixed sentiment is supported, but the issue is likely short text.

Trap 3: The posts are in a language that is not supported by the sentiment…

The service supports many languages; this is not the likely cause.

Study all Implement natural language processing solutions common traps →

A
The service does not support sentiment analysis for social media language.
Why it fails: The service supports social media language.
B
The posts are too short, causing the sentiment detection to default to neutral.
Short texts provide insufficient context for accurate sentiment detection.
C
The service is not configured to detect mixed sentiment.
Why it fails: Mixed sentiment is supported, but the issue is likely short text.
D
The posts are in a language that is not supported by the sentiment analysis API.
Why it fails: The service supports many languages; this is not the likely cause.

Full breakdown with real-world context →

Question 3hardmulti select

Read the full Implement natural language processing solutions explanation →

Which THREE components are required to build a custom named entity recognition (NER) model in Azure Cognitive Service for Language?

Trap 1: A QnA Maker knowledge base for entity definitions.

QnA Maker is for question answering, not NER.

Trap 2: A LUIS application to handle entity extraction.

LUIS is a separate service for conversational NLU.

Study all Implement natural language processing solutions common traps →

A
A QnA Maker knowledge base for entity definitions.
Why it fails: QnA Maker is for question answering, not NER.
B
A LUIS application to handle entity extraction.
Why it fails: LUIS is a separate service for conversational NLU.
C
A set of labeled documents for training.
Labeled data is essential for training.
D
A language service project with a custom NER schema.
The project defines entity types and tags.
E
A deployed endpoint for the custom NER model.
The model must be deployed for inferencing.

Full breakdown with real-world context →

Question 4hardmultiple choice

Read the full Implement natural language processing solutions explanation →

A hospital uses Azure Cognitive Service for Language to extract medical entities from clinical notes. The extraction accuracy for medication names and dosages is low. The engineer needs to improve performance without adding new training data. Which solution should the engineer implement?

Trap 1: Add more training data with annotated entities.

The question states without adding new training data.

Trap 2: Retrain the Text Analytics for Health model with additional labeled…

The prebuilt model cannot be retrained.

Trap 3: Increase the confidence threshold for entity extraction.

This may improve precision but lower recall.

Study all Implement natural language processing solutions common traps →

A
Add more training data with annotated entities.
Why it fails: The question states without adding new training data.
B
Use custom entity recognition with a prebuilt healthcare entity component.
This combines custom and prebuilt entities to improve accuracy.
C
Retrain the Text Analytics for Health model with additional labeled data.
Why it fails: The prebuilt model cannot be retrained.
D
Increase the confidence threshold for entity extraction.
Why it fails: This may improve precision but lower recall.

Full breakdown with real-world context →

Question 5hardmultiple choice

Read the full Implement natural language processing solutions explanation →

You are analyzing a document using Azure Cognitive Service for Language named entity recognition. The exhibit shows a partial JSON response for entity extraction. The engineer notices that 'Jane Smith' has a low confidence score of 0.45. Which action should the engineer take to improve the confidence score for similar entities?

Exhibit

Refer to the exhibit.
{
  "version": "2.0",
  "analysis": {
    "entities": [
      {
        "category": "Person",
        "text": "John Doe",
        "offset": 0,
        "length": 8,
        "confidenceScore": 0.99
      },
      {
        "category": "Person",
        "text": "Jane Smith",
        "offset": 20,
        "length": 10,
        "confidenceScore": 0.45
      }
    ]
  }
}

Trap 1: Retrain the entity recognition model with labeled examples of 'Jane…

The pre-built NER model is not retrainable; custom training requires a separate Custom NER project with its own labelled dataset and deployment, and does not alter the pre-built model's scores. It is tempting because custom NER is the correct choice when domain-specific entity types must be recognised.

Trap 2: Use a different language detection model to improve entity…

Language detection selects the language model for text; it does not influence entity extraction confidence, and the document's language is already identified. It is tempting because swapping detection models is correct when mixed-language or ambiguous input is being misclassified before analysis.

Trap 3: Decrease the confidence threshold to 0.3 to include 'Jane Smith' as…

Lowering the threshold changes which results are surfaced, not the model's computed probability, so the underlying 0.45 score is unchanged. It is tempting because threshold tuning is the correct lever when filtering pre-built model output for recall in a downstream application.

Study all Implement natural language processing solutions common traps →

A
Retrain the entity recognition model with labeled examples of 'Jane Smith'.
Why it fails: The pre-built NER model is not retrainable; custom training requires a separate Custom NER project with its own labelled dataset and deployment, and does not alter the pre-built model's scores. It is tempting because custom NER is the correct choice when domain-specific entity types must be recognised.
B
Provide more context around the entity in the input text, such as titles or roles.
Named entity recognition infers entity type from surrounding linguistic cues, so sparse text yields weak signals and low confidence. Adding titles or roles supplies disambiguating context that raises the model's certainty for similar entities, directly addressing the 0.45 score.
C
Use a different language detection model to improve entity recognition.
Why it fails: Language detection selects the language model for text; it does not influence entity extraction confidence, and the document's language is already identified. It is tempting because swapping detection models is correct when mixed-language or ambiguous input is being misclassified before analysis.
D
Decrease the confidence threshold to 0.3 to include 'Jane Smith' as a valid entity.
Why it fails: Lowering the threshold changes which results are surfaced, not the model's computed probability, so the underlying 0.45 score is unchanged. It is tempting because threshold tuning is the correct lever when filtering pre-built model output for recall in a downstream application.

Full breakdown with real-world context →

Question 6hardmultiple choice

Read the full Implement natural language processing solutions explanation →

A large retail company deploys a custom text classification model using Azure Cognitive Service for Language to categorize customer support tickets into 'Billing', 'Technical', and 'General' categories. The model is trained on 10,000 labeled tickets from the past year. After deployment, the model performs well on new tickets but shows a significant drop in accuracy for tickets submitted during holiday seasons, where the volume of billing issues spikes. The engineering team suspects concept drift. They need to maintain high accuracy without manual retraining every season. Which action should the engineer take?

Trap 1: Add subcategories to the classification schema to better capture…

More categories increase complexity, not address drift.

Trap 2: Increase the number of training examples per category to 15,000…

More data from the same distribution does not fix drift.

Trap 3: Schedule monthly retraining using the original 10,000 tickets plus…

Monthly retraining may not be frequent enough for seasonal spikes.

Study all Implement natural language processing solutions common traps →

A
Add subcategories to the classification schema to better capture seasonal patterns.
Why it fails: More categories increase complexity, not address drift.
B
Enable active learning on the model and set up a human review loop for low-confidence predictions.
Active learning continuously improves the model with new data.
C
Increase the number of training examples per category to 15,000 each.
Why it fails: More data from the same distribution does not fix drift.
D
Schedule monthly retraining using the original 10,000 tickets plus the new tickets.
Why it fails: Monthly retraining may not be frequent enough for seasonal spikes.

Full breakdown with real-world context →

Question 7hardmultiple choice

Read the full Implement natural language processing solutions explanation →

A company is building a chatbot using Azure Language Service and wants to ensure that the chatbot can understand user intents and extract entities from user utterances. The chatbot must be able to handle multiple intents in a single utterance and must support pre-built entities such as numbers and dates. Which action should the developer take to configure the Language service accordingly?

Trap 1: Upgrade the Language service tier from Standard to Custom.

The Standard tier already supports multiple intents and pre-built entities; upgrading is unnecessary.

Trap 2: Configure the project to use Orchestration workflow.

Orchestration workflow is used to route utterances to different projects, not to enable multiple intents within a single project.

Trap 3: Set the project language to 'Multilingual' to enable entity…

The multilingual option is for supporting multiple languages, not for enabling multiple intents or pre-built entities.

Study all Implement natural language processing solutions common traps →

A
Enable the 'Multiple intents' setting in the Language service project.
Conversational language understanding projects process one intent per utterance by default. Enabling the multiple intents setting lets the model return several intents from a single utterance, satisfying the requirement to handle multiple intents while pre-built entities cover numbers and dates.
B
Upgrade the Language service tier from Standard to Custom.
Why it fails: The Standard tier already supports multiple intents and pre-built entities; upgrading is unnecessary.
C
Configure the project to use Orchestration workflow.
Why it fails: Orchestration workflow is used to route utterances to different projects, not to enable multiple intents within a single project.
D
Set the project language to 'Multilingual' to enable entity recognition.
Why it fails: The multilingual option is for supporting multiple languages, not for enabling multiple intents or pre-built entities.
E
Define list entities for numbers and dates.
Why it fails: Pre-built entities like numbers and dates are automatically recognized without needing to define them as list entities.

Full breakdown with real-world context →

Question 8easymultiple choice

Read the full Implement natural language processing solutions explanation →

A healthcare company is developing a solution to analyze patient feedback using Azure AI Language. The solution must extract key phrases, detect sentiment, and identify personally identifiable information (PII) such as patient names and medical record numbers from unstructured text. The company has strict compliance requirements: all text processing must occur within the United States region, and no data may leave the Azure geography. The development team has provisioned a Language resource in the East US region and has been testing the solution. During testing, the team notices that the PII detection feature is returning results, but the key phrase extraction and sentiment analysis are failing with a 403 error. The error message indicates that the resource is not allowed to access these features. The team has verified that the resource is in the S0 tier. What should the team do to resolve the issue?

Trap 1: Switch to the free tier to access all features.

The free tier has limited features and may not support all required features; also the issue is regional, not tier-based.

Trap 2: Upgrade the Language resource to the S1 tier to enable all features.

The S0 tier already supports key phrase extraction and sentiment analysis; the issue is not the tier.

Trap 3: Regenerate the API key and update the application code.

Regenerating the key does not affect feature availability.

Study all Implement natural language processing solutions common traps →

A
Switch to the free tier to access all features.
Why it fails: The free tier has limited features and may not support all required features; also the issue is regional, not tier-based.
B
Upgrade the Language resource to the S1 tier to enable all features.
Why it fails: The S0 tier already supports key phrase extraction and sentiment analysis; the issue is not the tier.
C
Verify regional availability of the features and recreate the Language resource in a supported region if necessary.
Key phrase extraction and sentiment analysis are not offered in every Azure region, whereas PII detection is more widely deployed. The 403 indicates the East US resource cannot access those features, so confirming regional support and recreating the Language resource in a supported region resolves it.
D
Regenerate the API key and update the application code.
Why it fails: Regenerating the key does not affect feature availability.

Full breakdown with real-world context →

Question 9mediumdrag order

Read the full Implement natural language processing solutions explanation →

Drag and drop the steps to deploy a custom language model using Azure AI Language into the correct order.

Drag or tap steps into the slots.

Steps

Order

1Step 1

2Step 2

3Step 3

4Step 4

5Step 5

Trap 1: Step 1: Create the Azure AI Language resource, Step 2: Have labeled…

This is incorrect because you should have labeled data ready before creating the resource to ensure the resource is configured appropriately for your data.

Trap 2: Step 1: Have labeled data ready, Step 2: Create the Azure AI…

This is incorrect because you cannot evaluate a model before training it; evaluation must follow training.

Trap 3: Step 1: Have labeled data ready, Step 2: Create the Azure AI…

This is incorrect because you must evaluate the model before deploying it to ensure it meets performance requirements.

Study all Implement natural language processing solutions common traps →

A
Step 1: Have labeled data ready, Step 2: Create the Azure AI Language resource, Step 3: Train the custom language model, Step 4: Evaluate the model performance, Step 5: Deploy the model to an endpoint
This is the correct order because you need data before creating the resource, then train, evaluate, and finally deploy.
B
Step 1: Create the Azure AI Language resource, Step 2: Have labeled data ready, Step 3: Train the custom language model, Step 4: Evaluate the model performance, Step 5: Deploy the model to an endpoint
Why it fails: This is incorrect because you should have labeled data ready before creating the resource to ensure the resource is configured appropriately for your data.
C
Step 1: Have labeled data ready, Step 2: Create the Azure AI Language resource, Step 3: Evaluate the model performance, Step 4: Train the custom language model, Step 5: Deploy the model to an endpoint
Why it fails: This is incorrect because you cannot evaluate a model before training it; evaluation must follow training.
D
Step 1: Have labeled data ready, Step 2: Create the Azure AI Language resource, Step 3: Train the custom language model, Step 4: Deploy the model to an endpoint, Step 5: Evaluate the model performance
Why it fails: This is incorrect because you must evaluate the model before deploying it to ensure it meets performance requirements.

Full breakdown with real-world context →

Question 10mediummatching

Read the full Implement natural language processing solutions explanation →

Match each Azure AI term to its definition.

Drag a concept onto its matching description — or click a concept then click the description.

Concepts

Matches

Language Understanding Intelligent Service

Service to create a question and answer bot

Service to build custom image classifiers

Convert spoken language to text

Extract insights from text like key phrases

Trap 1: Azure Cognitive Services: A service for building and training…

Incorrect — this describes Azure Machine Learning, not Cognitive Services.

Trap 2: Azure Machine Learning: A collection of pre-built AI APIs for…

Incorrect — this defines Azure Cognitive Services, not Azure Machine Learning.

Trap 3: Azure Bot Service: A service that provides pre-built AI models for…

Incorrect — this refers to Cognitive Services (e.g., Language Understanding), not Bot Service.

Study all Implement natural language processing solutions common traps →

A
Azure Cognitive Services: A collection of AI services and APIs that help you build intelligent applications without needing direct AI or data science skills.
This is the correct definition of Azure Cognitive Services, which provides pre-built AI models via APIs.
B
Azure Machine Learning: A cloud service for end-to-end machine learning lifecycle management, including training, deployment, and management.
This correctly describes Azure Machine Learning as a platform for building, training, and managing ML models.
C
Azure Bot Service: A service for building, testing, deploying, and managing intelligent bots.
This is the accurate definition of Azure Bot Service, which facilitates bot development and management.
D
Azure Cognitive Services: A service for building and training machine learning models.
Why it fails: Incorrect — this describes Azure Machine Learning, not Cognitive Services.
E
Azure Machine Learning: A collection of pre-built AI APIs for vision, speech, language, etc.
Why it fails: Incorrect — this defines Azure Cognitive Services, not Azure Machine Learning.
F
Azure Bot Service: A service that provides pre-built AI models for natural language processing.
Why it fails: Incorrect — this refers to Cognitive Services (e.g., Language Understanding), not Bot Service.

Full breakdown with real-world context →

Question 11mediummultiple choice

Read the full Implement natural language processing solutions explanation →

You are building a multilingual support chatbot using Azure AI Language. The chatbot must understand user queries in English, Spanish, and French, and respond in the same language. The solution should minimize latency and cost. What is the recommended approach?

Trap 1: Use Azure AI Translator to translate all queries to English,…

Translator adds latency and cost, and may lose context.

Trap 2: Create separate Azure AI Language projects for each language and…

Multiple projects increase management overhead and cost.

Trap 3: Use a single English-only project and rely on Azure AI Translator…

English-only with translation adds latency and may lose context.

Study all Implement natural language processing solutions common traps →

A
Use Azure AI Translator to translate all queries to English, process with an English-only project, then translate responses back.
Why it fails: Translator adds latency and cost, and may lose context.
B
Use a single Azure AI Language project with multilingual support enabled.
A single multilingual project shares one model and endpoint across English, Spanish and French, avoiding the latency and per-resource cost of deploying three separate projects. Language detection and same-language replies are handled natively, meeting the minimisation constraint.
C
Create separate Azure AI Language projects for each language and route queries based on detected language.
Why it fails: Multiple projects increase management overhead and cost.
D
Use a single English-only project and rely on Azure AI Translator for all non-English queries.
Why it fails: English-only with translation adds latency and may lose context.

Full breakdown with real-world context →

Question 12mediummultiple choice

Read the full Implement natural language processing solutions explanation →

You are developing a conversational agent using Microsoft Copilot Studio that must handle complex multi-turn conversations. The agent needs to maintain context across multiple user inputs. Which feature should you use?

Trap 1: Use actions to call external APIs for context.

Actions are for external integration, not conversation management.

Trap 2: Use global variables to store conversation state.

Variables store data, but slots provide structured conversation flow.

Trap 3: Use entities to capture user input.

Entities capture data, but not conversation flow.

Study all Implement natural language processing solutions common traps →

A
Use actions to call external APIs for context.
Why it fails: Actions are for external integration, not conversation management.
B
Use topics with slots to collect information across turns.
Slots within topics persist values gathered across separate conversational turns, letting the agent retain context until every required slot is filled. This directly satisfies the multi-turn context requirement, unlike single-response alternatives that discard state between user inputs.
C
Use global variables to store conversation state.
Why it fails: Variables store data, but slots provide structured conversation flow.
D
Use entities to capture user input.
Why it fails: Entities capture data, but not conversation flow.

Full breakdown with real-world context →

Question 13mediummultiple choice

Read the full Implement natural language processing solutions explanation →

Refer to the exhibit. You submit this request to the Azure AI Language service. What is the expected response?

Exhibit

Refer to the exhibit.
```json
{
  "analysisInput": {
    "documents": [
      {
        "id": "1",
        "text": "The quick brown fox jumps over the lazy dog.",
        "language": "en"
      }
    ]
  },
  "tasks": [
    {
      "taskName": "EntityRecognition",
      "kind": "EntityRecognition",
      "parameters": {
        "modelVersion": "2022-10-01",
        "stringIndexType": "TextElement_V8"
      }
    },
    {
      "taskName": "KeyPhraseExtraction",
      "kind": "KeyPhraseExtraction",
      "parameters": {
        "modelVersion": "2024-01-01"
      }
    }
  ]
}
```

Trap 1: An error because the request format is invalid.

The payload is well-formed JSON matching the analyze-text schema, so the service processes it and returns results rather than a 400. Expecting an error is tempting when a request looks unusual, yet malformed input would require a schema violation such as an unsupported task name or missing documents array.

Trap 2: Only key phrase extraction results.

The request targets the /analyze-text endpoint, which returns every requested task in one payload; key phrases alone appear only when that is the sole task specified. Selecting key phrase extraction is tempting because it is a valid Azure AI Language feature, but it answers a single-task request, not this multi-task one.

Trap 3: Only entity recognition results.

Entity recognition is only one of several tasks submitted, so the response contains results for every task, not entities alone. Restricting to entities is tempting because entity recognition is a core Azure AI Language capability, but it would be the whole answer only if the request specified that single task.

Study all Implement natural language processing solutions common traps →

A
An error because the request format is invalid.
Why it fails: The payload is well-formed JSON matching the analyze-text schema, so the service processes it and returns results rather than a 400. Expecting an error is tempting when a request looks unusual, yet malformed input would require a schema violation such as an unsupported task name or missing documents array.
B
Only key phrase extraction results.
Why it fails: The request targets the /analyze-text endpoint, which returns every requested task in one payload; key phrases alone appear only when that is the sole task specified. Selecting key phrase extraction is tempting because it is a valid Azure AI Language feature, but it answers a single-task request, not this multi-task one.
C
Both entity recognition and key phrase extraction results.
The Azure AI Language analyse endpoint accepts a documents array and returns a combined JSON payload containing both the entities collection and the keyPhrases collection, so a single request yields both result sets rather than one or the other.
D
Only entity recognition results.
Why it fails: Entity recognition is only one of several tasks submitted, so the response contains results for every task, not entities alone. Restricting to entities is tempting because entity recognition is a core Azure AI Language capability, but it would be the whole answer only if the request specified that single task.

Full breakdown with real-world context →

Question 14mediummultiple choice

Read the full Implement natural language processing solutions explanation →

A company deploys a custom question answering project in Azure AI Language. Users report that the bot sometimes returns irrelevant answers. The knowledge base contains hundreds of QnA pairs. You need to improve answer relevance without retraining the model. What should you do?

Trap 1: Increase the confidence score threshold in the project settings.

This may cause many valid queries to be unanswered.

Trap 2: Reduce the number of QnA pairs to decrease ambiguity.

This may remove valid content and reduce coverage.

Trap 3: Enable active learning to let the bot suggest new questions based…

Active learning requires user feedback over time; it does not immediately improve relevance.

Study all Implement natural language processing solutions common traps →

A
Increase the confidence score threshold in the project settings.
Why it fails: This may cause many valid queries to be unanswered.
B
Reduce the number of QnA pairs to decrease ambiguity.
Why it fails: This may remove valid content and reduce coverage.
C
Add alternate phrases to existing QnA pairs for common user queries.
Alternate phrases expand each QnA pair's surface-form coverage, letting the ranker match paraphrased user queries to the intended answer without retraining. This raises relevance by improving lexical and semantic overlap between incoming questions and stored question variants.
D
Enable active learning to let the bot suggest new questions based on user queries.
Why it fails: Active learning requires user feedback over time; it does not immediately improve relevance.

Full breakdown with real-world context →

Question 15easymultiple choice

Read the full Implement natural language processing solutions explanation →

You are building a chatbot that uses Azure AI Language to extract intents and entities from user utterances. The bot must recognize custom entities like product names that are not in the default model. Which feature should you use?

Trap 1: Prebuilt entity recognition component.

Prebuilt entities cover general categories such as datetime, numbers and email, not bespoke product names absent from the default model. It is tempting because prebuilt recognition needs no training data, and would be correct when extracting common, domain-independent entity types rather than organisation-specific vocabulary.

Trap 2: Key phrase extraction.

Key phrase extraction is an independent text-analytics capability that returns salient phrases; it does not feed intents or entities into a CLU project. It is tempting because it surfaces important terms, and would be correct for summarising or indexing documents rather than resolving utterance entities during conversation.

Trap 3: List entity in a conversational language understanding (CLU)…

A list entity matches exact, closed sets of terms, so it cannot generalise to unseen product names or learn from labelled utterances. It is tempting because list entities are simple to author, and would be correct for a fixed, known catalogue of values rather than custom entities requiring trained recognition.

Study all Implement natural language processing solutions common traps →

A
Prebuilt entity recognition component.
Why it fails: Prebuilt entities cover general categories such as datetime, numbers and email, not bespoke product names absent from the default model. It is tempting because prebuilt recognition needs no training data, and would be correct when extracting common, domain-independent entity types rather than organisation-specific vocabulary.
B
Key phrase extraction.
Why it fails: Key phrase extraction is an independent text-analytics capability that returns salient phrases; it does not feed intents or entities into a CLU project. It is tempting because it surfaces important terms, and would be correct for summarising or indexing documents rather than resolving utterance entities during conversation.
C
Custom named entity recognition (NER) component.
Custom named entity recognition trains a model on your labelled product-name utterances, so it extracts entities absent from the prebuilt model. This directly satisfies the stem's constraint: recognising custom entities such as product names. Azure AI Language's custom NER component supports this within the same resource, unlike prebuilt NER, which only returns fixed entity categories.
D
List entity in a conversational language understanding (CLU) project.
Why it fails: A list entity matches exact, closed sets of terms, so it cannot generalise to unseen product names or learn from labelled utterances. It is tempting because list entities are simple to author, and would be correct for a fixed, known catalogue of values rather than custom entities requiring trained recognition.

Full breakdown with real-world context →

Question 16hardmultiple choice

Read the full Implement natural language processing solutions explanation →

A company uses Azure AI Language for sentiment analysis on customer feedback. They notice that the sentiment scores for mixed reviews are often neutral when they should be slightly positive. They need to improve the accuracy for these mixed reviews without labeling new data. Which approach should you recommend?

Trap 1: Adjust the sentiment threshold to require a higher confidence for…

This may cause genuine neutral reviews to be misclassified.

Trap 2: Use a prebuilt sentiment analysis model with a different language.

Changing language is not relevant to domain-specific accuracy.

Trap 3: Create and train a custom sentiment analysis model using labeled…

A custom model trained on domain data will better handle mixed reviews.

Study all Implement natural language processing solutions common traps →

A
Adjust the sentiment threshold to require a higher confidence for neutral.
Why it fails: This may cause genuine neutral reviews to be misclassified.
B
Enable opinion mining to capture aspects and opinions separately.
Opinion mining adds detail but does not improve the base sentiment model.
C
Use a prebuilt sentiment analysis model with a different language.
Why it fails: Changing language is not relevant to domain-specific accuracy.
D
Create and train a custom sentiment analysis model using labeled feedback data.
Why it fails: A custom model trained on domain data will better handle mixed reviews.

Full breakdown with real-world context →

Question 17hardmultiple choice

Read the full Implement natural language processing solutions explanation →

A company uses Azure AI Language to analyze customer call transcripts. They need to identify specific entities such as product names and issue types. The prebuilt entity recognition does not cover their custom entities. Which approach should they take to extract both standard and custom entities from the transcripts?

Trap 1: Use the prebuilt entity recognition API only and map standard…

Prebuilt entities do not recognize custom entities.

Trap 2: Use a single conversational language understanding (CLU) project…

CLU projects define entities manually, not using prebuilt models.

Trap 3: Use a custom text classification model to classify the transcript…

Text classification does not extract entities.

Study all Implement natural language processing solutions common traps →

A
Use the prebuilt entity recognition API only and map standard entities to custom categories.
Why it fails: Prebuilt entities do not recognize custom entities.
B
Use a single conversational language understanding (CLU) project with entities defined for both standard and custom entities.
Why it fails: CLU projects define entities manually, not using prebuilt models.
C
Use a custom text classification model to classify the transcript and then extract entities from the classified output.
Why it fails: Text classification does not extract entities.
D
Use a custom named entity recognition (NER) model for custom entities and call the prebuilt entity recognition API separately for standard entities.
Custom NER handles the bespoke entity types, but it cannot emit the prebuilt categories. Calling the prebuilt entity recognition endpoint alongside it lets both standard and custom entities be extracted from the same transcripts, satisfying the stem's combined requirement.

Full breakdown with real-world context →

Question 18hardmultiple choice

Read the full Implement natural language processing solutions explanation →

A company uses Azure AI Language's conversational language understanding (CLU) to build a customer support bot. They want to integrate the bot with Microsoft Teams and need to ensure that user authentication is handled by Microsoft Entra ID. However, users report that the bot sometimes fails to respond when they are not signed into Microsoft Entra ID. What is the most likely cause?

Trap 1: The CLU model requires a Microsoft Entra ID token for prediction.

The model does not require Microsoft Entra ID for prediction.

Trap 2: The CLU project endpoint is not configured to accept anonymous…

The endpoint can accept anonymous requests; authentication is at the bot level.

Trap 3: The bot is not registered in Microsoft Entra ID.

The bot must be registered, but the issue is user authentication.

Study all Implement natural language processing solutions common traps →

A
The CLU model requires a Microsoft Entra ID token for prediction.
Why it fails: The model does not require Microsoft Entra ID for prediction.
B
The CLU project endpoint is not configured to accept anonymous requests.
Why it fails: The endpoint can accept anonymous requests; authentication is at the bot level.
C
The bot is not registered in Microsoft Entra ID.
Why it fails: The bot must be registered, but the issue is user authentication.
D
The bot's authentication settings in Azure Bot Service require Microsoft Entra ID, but the bot is not passing the token correctly.
Azure Bot Service authentication configured for Microsoft Entra ID requires a valid token on each request. If the bot fails to pass or refresh that token correctly, unauthenticated users trigger failures, matching the reported symptom of non-responsiveness when users are not signed in.

Full breakdown with real-world context →

Question 19mediummulti select

Read the full Implement natural language processing solutions explanation →

Which TWO actions should you take to improve the performance of a custom named entity recognition (NER) model in Azure AI Language?

Trap 1: Increase the training time of the model.

Training duration is not a tunable lever for NER quality; Azure AI Language stops when convergence criteria are met, so extra time changes nothing. It tempts because longer training helps some deep-learning models, and would be right if the model were underfitting due to too few epochs on a small labelled set.

Trap 2: Reduce the number of entity types to simplify the model.

Fewer entity types reduces the model's ability to distinguish domain-specific categories, lowering recall on the entities you need. Custom NER benefits from clearly separated, well-labelled entity types. Reducing scope suits a proof of concept, not performance improvement; adding labelled examples per entity is the effective action.

Trap 3: Use only prebuilt entity types to avoid training from scratch.

Restricting to prebuilt entity types abandons custom training entirely, so the custom NER model cannot improve. Prebuilt entities suit scenarios needing general categories like person or location without labelled domain data. Adding labelled utterances and increasing training data volume are the actual levers for custom model accuracy.

Study all Implement natural language processing solutions common traps →

A
Use a balanced dataset with similar numbers of examples for each entity.
Class imbalance biases the model toward over-represented entities, degrading recall on rare ones. Training with comparable example counts per entity ensures the model learns each entity's patterns rather than favouring the majority label, directly improving overall extraction performance.
B
Increase the training time of the model.
Why it fails: Training duration is not a tunable lever for NER quality; Azure AI Language stops when convergence criteria are met, so extra time changes nothing. It tempts because longer training helps some deep-learning models, and would be right if the model were underfitting due to too few epochs on a small labelled set.
C
Reduce the number of entity types to simplify the model.
Why it fails: Fewer entity types reduces the model's ability to distinguish domain-specific categories, lowering recall on the entities you need. Custom NER benefits from clearly separated, well-labelled entity types. Reducing scope suits a proof of concept, not performance improvement; adding labelled examples per entity is the effective action.
D
Label more examples with entity annotations.
Additional annotated utterances expand the training set, letting the model learn entity boundaries and context it currently misses. This directly addresses the stem's performance constraint by reducing misclassifications on underrepresented entity types, provided the new labels are balanced and consistent with existing annotation guidelines.
E
Use only prebuilt entity types to avoid training from scratch.
Why it fails: Restricting to prebuilt entity types abandons custom training entirely, so the custom NER model cannot improve. Prebuilt entities suit scenarios needing general categories like person or location without labelled domain data. Adding labelled utterances and increasing training data volume are the actual levers for custom model accuracy.

Full breakdown with real-world context →

Question 20hardmulti select

Read the full Implement natural language processing solutions explanation →

Which THREE components are required to deploy a bot using Azure AI Language's conversational language understanding (CLU) and Azure Bot Service?

Trap 1: A QnA Maker resource for FAQs.

Not required; CLU can handle FAQs.

Trap 2: A Translator resource for multilingual support.

Not required; CLU can handle multiple languages.

Study all Implement natural language processing solutions common traps →

A
An Azure Bot Service bot with the CLU model as a skill.
The bot itself is mandatory: an Azure Bot Service resource hosts the conversational endpoint and channels, and it must be wired to call the CLU model as a recognised skill so utterances reach the language understanding service for intent and entity resolution.
B
A CLU project and a trained model.
CLU requires a project containing your intents, entities and utterances, plus a successfully trained model that the runtime can query. Without a trained model, the bot has no published prediction endpoint to send utterances to, so deployment cannot function.
C
A QnA Maker resource for FAQs.
Why it fails: Not required; CLU can handle FAQs.
D
A Translator resource for multilingual support.
Why it fails: Not required; CLU can handle multiple languages.
E
A Microsoft Entra ID app registration for the bot.
The bot's Microsoft Entra ID app registration supplies the identity, application ID and secret or certificate used for authentication between Azure Bot Service and the bot's messaging endpoint, which is required for the channel connection to be authorised.

Full breakdown with real-world context →

Continue with 20-question session →

Free account

Track your progress over time

Create a free account to save your results and see which topics improve across sessions.

Focused Implement natural language processing solutions sessions

Start a Implement natural language processing solutions only practice session

Every question in these sessions is drawn from the Implement natural language processing solutions domain — nothing else.

10 questions 20 questions 30 questions 50 questions

Browse all Implement natural language processing solutions questions →Mixed AI-102 session

Frequently asked questions

What does the AI-102 exam test about Implement natural language processing solutions?: Be able to select the correct Azure AI Language or Speech feature for a scenario, configure multilingual intent recognition, and secure access to the resource. The most important thing is matching the service and its language or network configuration to the stated requirement.
How should I use these practice questions?: Select your answer before revealing the explanation. Then read why each option is right or wrong — this active recall approach builds retention far faster than re-reading notes.
Can I practise just Implement natural language processing solutions questions in a focused session?: Yes — the session launcher on this page draws every question from the Implement natural language processing solutions domain. Use a 10-question session first to gauge your baseline, then move to 20 or 30 once the weak spots are clear.
Where can I practise other AI-102 topics?: Use the topic links above to move to related areas, or go back to the AI-102 question bank to see all topics.
Are these real exam questions or dumps?: These are original practice questions written to test the same concepts the AI-102 exam covers. They are not copied from any real exam or dump site.

Implement natural language processing solutions only

10 questions 20 questions 30 questions 50 questions

Mixed AI-102 session

Track your progress

A free account saves results across sessions and highlights which topics need work.

Study resources

All AI-102 questions Implement natural language processing solutions domain overview AI-102 exam guide

Exam traps to avoid

▸Assuming one LUIS or CLU model automatically handles all languages; multilingual support requires explicit configuration or separate language models.
▸Confusing batch transcription with real-time speech-to-text when latency requirements are stated.
▸Forgetting that allowing on-premises access requires configuring firewall, virtual network, or private endpoint settings, not just copying a key.

Implement natural language processing solutions practice questions

What to know about Implement natural language processing solutions

Why Implement natural language processing solutions questions are commonly missed

Common Implement natural language processing solutions exam traps

Implement natural language processing solutions questions

A company is building a chatbot using Azure Cognitive Service for Language. They need to ensure that user utterances are correctly mapped to the appropriate intent in a custom question answering project. What should they configure?

A development team is using the Azure Cognitive Service for Language to perform sentiment analysis on social media posts. They notice that the returned sentiment scores are often neutral for posts that are clearly positive or negative. What is the most likely reason?

Which THREE components are required to build a custom named entity recognition (NER) model in Azure Cognitive Service for Language?

A hospital uses Azure Cognitive Service for Language to extract medical entities from clinical notes. The extraction accuracy for medication names and dosages is low. The engineer needs to improve performance without adding new training data. Which solution should the engineer implement?

Exhibit

Drag and drop the steps to deploy a custom language model using Azure AI Language into the correct order.

Match each Azure AI term to its definition.

You are building a multilingual support chatbot using Azure AI Language. The chatbot must understand user queries in English, Spanish, and French, and respond in the same language. The solution should minimize latency and cost. What is the recommended approach?

You are developing a conversational agent using Microsoft Copilot Studio that must handle complex multi-turn conversations. The agent needs to maintain context across multiple user inputs. Which feature should you use?

Refer to the exhibit. You submit this request to the Azure AI Language service. What is the expected response?

Exhibit

A company deploys a custom question answering project in Azure AI Language. Users report that the bot sometimes returns irrelevant answers. The knowledge base contains hundreds of QnA pairs. You need to improve answer relevance without retraining the model. What should you do?

You are building a chatbot that uses Azure AI Language to extract intents and entities from user utterances. The bot must recognize custom entities like product names that are not in the default model. Which feature should you use?

Which TWO actions should you take to improve the performance of a custom named entity recognition (NER) model in Azure AI Language?

Which THREE components are required to deploy a bot using Azure AI Language's conversational language understanding (CLU) and Azure Bot Service?

Track your progress over time

Start a Implement natural language processing solutions only practice session

Related AI-102 topic practice pages

Implement computer vision solutions practice questions

Implement knowledge mining and information extraction solutions practice questions

Implement generative AI solutions practice questions

Plan and manage an Azure AI solution practice questions

Implement agentic AI solutions practice questions

Implement an agentic solution practice questions

Implement knowledge mining and document intelligence solutions practice questions