AI-102 Implement natural language processing solutions Practice Test 6 — 15 Questions

Question 1

You are building a solution to analyze customer feedback from multiple sources: emails, chat logs, and survey responses. You need to detect the overall sentiment trend over time and identify the most frequently mentioned topics. The solution must also allow the business analyst to ask natural language questions about the data (e.g., 'Show me complaints about shipping in the last month'). You have all data in Azure Blob Storage. You need to implement a solution with minimal custom code. Which combination of Azure services should you use?

Choose an answer to begin — your selection is scored in the full session.

Accepted Answer

Use Azure AI Language to extract sentiment and key phrases, then index the data in Azure Cognitive Search with semantic search enabled; use the search's built-in features for trend analysis and natural language queries.. Azure AI Language provides built-in sentiment analysis and key phrase extraction, and Azure Cognitive Search with semantic search enables indexing the extracted data for trend analysis and natural language queries without custom code. This combination directly meets the requirements of detecting sentiment trends, identifying topics, and allowing natural language questions, all with minimal custom development.

Answer

Use Azure OpenAI Service to analyze sentiment and generate summaries, and store results in a Cosmos DB for querying.

Answer

Use Azure AI Language to perform sentiment analysis and key phrase extraction, then load the results into Power BI for trend analysis and natural language Q&A.

Answer

Use Azure AI Language's custom question answering to create a knowledge base from the feedback and allow natural language queries.

Question 2

Your company uses Azure AI Language to process legal documents. You have built a custom NER model to extract parties, dates, and obligations. The model performs well on English documents but now you need to support French and German documents. You have no labeled data in those languages. You want to use the existing English model as a starting point. The solution must be cost-effective and avoid manual labeling as much as possible. You also need to ensure that the model can be retrained quickly when new document types are added. Which approach should you take?

Accepted Answer

Enable the multilingual option in the custom NER project and retrain the model using the existing English labeled data.. Azure AI Language custom NER supports a multilingual option that lets you train a single model using labeled data in one language and have it apply across multiple languages. By enabling this option and retraining with the existing English labeled data, the model can extract the same custom entity types from French and German documents without requiring new labeled data. This is the most cost-effective path because it reuses existing annotations and avoids translation pipelines or duplicate model maintenance.

Answer

Use the prebuilt NER model for French and German to extract entities like dates and organizations, and then map them to your custom entity types.

Answer

Use the English custom NER model as-is and translate the French and German documents to English before processing.

Answer

Build separate custom NER models for French and German by translating the English labeled data using Azure AI Translator and training new models.

Question 3

A healthcare organization is deploying a solution using Azure AI Language to extract medical entities from clinical notes. The solution must comply with HIPAA and support the following requirements: extract medication names, dosages, and frequencies; identify patient conditions; and recognize negated terms (e.g., 'no sign of infection'). Which THREE Azure AI Language features should the organization use?

Accepted Answer

Prebuilt NER for Healthcare. Option B (Prebuilt NER for Healthcare) is correct because Azure AI Language's Text Analytics for Health model is purpose-built to extract medical entities such as medication names, dosages, frequencies, and conditions from clinical notes, satisfying the extraction requirements. Option D (Negation detection) is correct because Text Analytics for Health includes assertion detection that identifies negated terms like 'no sign of infection', which is explicitly required. Option E (Custom Named Entity Recognition) is correct because custom NER allows the organization to train a model on its own labeled clinical data to capture domain-specific entities and terminology not covered by the prebuilt model, supporting HIPAA-aligned customization. Option A (PII detection) is not correct because, while useful for de-identification, it does not extract medications, dosages, conditions, or negations. Option C (Prebuilt NER for Finance) is not correct because finance entities are irrelevant to clinical note processing.

Answer

PII detection

Answer

Prebuilt NER for Finance

Question 4

A legal firm is using Azure AI Language to analyze contracts. They need to extract key clauses, parties involved, and dates. The solution must be customizable to their specific contract types. Which TWO Azure AI Language features should they use?

Accepted Answer

Prebuilt NER for Legal. Prebuilt NER for Legal is correct because it provides out-of-the-box entity extraction tailored to legal documents, including parties, dates, and key clauses, without requiring custom training. This feature is specifically designed for legal use cases, making it ideal for a firm that needs to quickly extract standard legal entities from contracts.

Answer

Conversation summarization

Answer

Key phrase extraction

Answer

Entity linking

Question 5

Refer to the exhibit. You have a Custom Question Answering project configured with the JSON shown. When you test the project in Azure AI Language Studio, the query 'How many vacation days do I get?' returns no answer. What is the most likely cause?

Accepted Answer

The query is not phrased as an exact match to the trained question.. Custom Question Answering can be configured to require exact matching between the user query and the trained questions. In the exhibit, the JSON likely defines a QnA pair with a specific question, but the test query 'How many vacation days do I get?' does not exactly match the trained question (e.g., it might be slightly different wording). Since the project is set to exact match, the service returns no answer because the query is not an exact match. Option A correctly identifies this cause.

Answer

The language is set to English but the query uses informal language.

Answer

The answer field is empty in the JSON.

Answer

The confidence score threshold is set too high.

Question 6

Refer to the exhibit. You called the Key Phrase Extraction API on two documents. What is the total number of key phrases extracted?

Accepted Answer

5. The Key Phrase Extraction API returns a list of key phrases per document. In this case, Document 1 returns 'weather', 'Seattle', and 'rainy', while Document 2 returns 'weather' and 'Seattle'. The total number of unique key phrases across both documents is 3 ('weather', 'Seattle', 'rainy'), but the question asks for the total number of key phrases extracted, which counts each occurrence per document. That gives 3 + 2 = 5 key phrases. Option B is correct because the API outputs phrases per document, not deduplicated globally.

Answer

7

Answer

3

Answer

2

Question 7

Refer to the exhibit. You receive this error when calling an Azure Cognitive Services API. What is the most likely cause?

Accepted Answer

The subscription key is not valid for this Cognitive Services resource.. The error indicates that the subscription key provided is not associated with the Cognitive Services resource being accessed. In Azure Cognitive Services, each resource has a unique set of keys, and using a key from a different resource (even in the same region) will result in a 401 Unauthorized error with this specific message. Option D is correct because the key must match the resource's endpoint exactly.

Answer

The API endpoint is incorrect.

Answer

The subscription key has expired.

Answer

The subscription key is from a different region.

Question 8

Refer to the exhibit. A developer is debugging an Azure web app that uses Azure AD authentication. The app frequently fails to authenticate users. What is the most likely cause of the error?

Accepted Answer

The web app's outbound traffic is blocked by a firewall or VNet restriction.. The most likely cause of authentication failures is that the web app's outbound traffic is blocked by a firewall or VNet restriction. Microsoft Entra ID authentication requires the app to make outbound calls to Microsoft Entra ID endpoints for token validation, metadata retrieval, and sometimes user info. If outbound traffic is blocked, these calls fail, causing authentication errors. While expired JWT tokens, missing app registration, or incorrect client secret can cause failures, they are less likely if the app frequently fails intermittently, which points to network issues.

Answer

The JWT token has expired.

Answer

The web app is not registered in Azure AD.

Answer

The web app is using an incorrect client secret.

Question 9

Refer to the exhibit. A developer is training a Language Understanding (LUIS) model. When testing the phrase 'Show me the status of order 98765', the model returns the intent 'OrderStatus' with a low confidence score. What is the most likely reason?

Accepted Answer

The phrase 'Show me the status' was not seen during training.. LUIS models rely on labeled training utterances to learn intents and entities. If the phrase 'Show me the status' was not present in any training utterance for the 'OrderStatus' intent, the model lacks the patterns to recognize it, resulting in low confidence. The presence of an unlabeled entity might affect entity extraction but not intent confidence as directly.

Answer

The entity 'OrderNumber' should be a list entity.

Answer

The phrase contains an entity that was not labeled in training.

Answer

The model is overfitted to the training data.

Question 10

Refer to the exhibit. A developer tries to deploy a custom speech model but receives an error. What should the developer do to resolve the issue?

Accepted Answer

Specify a specific version number for the model.. The error states that 'latest' is not supported; a specific version must be specified.

Answer

Use a different model format.

Answer

Use the model version 'current' instead of 'latest'.

Answer

Re-train the model with more data.

Question 11

You are building a customer support application that must detect the language of incoming chat messages and return a confidence score for the detected language. You call the Azure AI Language service using the REST API. Which request should you use to obtain both the detected language and a confidence score?

Accepted Answer

POST to /language/:analyze-text?api-version=2023-04-01 with kind set to "LanguageDetection" and a documents array containing the text.. The Azure AI Language unified analyze-text endpoint supports LanguageDetection when kind is set to LanguageDetection in a POST body with a documents array. The response includes the detected language and a confidence score, matching the requirement. Other kinds such as SentimentAnalysis or older API versions do not provide the requested language and confidence output in the same way.

Answer

POST to /language/:analyze-text?api-version=2023-04-01 with kind set to "SentimentAnalysis" and include the text.

Answer

POST to /text/analytics/v3.1/languages with the text in the body and a header Ocp-Apim-Subscription-Key.

Answer

GET to /language/:analyze-text?api-version=2023-04-01 with the text in the query string and kind=LanguageDetection.

Question 12

You are deploying a custom named entity recognition (NER) model in Azure AI Language. The model is trained to extract 'ProductName' and 'OrderID' entities from customer support emails. After deployment, you notice that the model frequently misclassifies 'OrderID' as a general number. You need to improve the model's performance without retraining from scratch. What should you do?

Accepted Answer

Use the model's evaluation metrics to identify misclassified entities and add them as labeled examples in the training set, then retrain.. The correct approach is to use the model's evaluation metrics to identify specific misclassifications, add those examples as labeled data, and retrain the model. This targeted improvement addresses the root cause without starting from scratch. Other options either don't directly fix the misclassification or are not applicable to custom NER models in Azure AI Language.

Answer

Add more labeled examples of 'OrderID' entities in the training data and retrain the model.

Answer

Increase the model's confidence threshold for entity recognition to reduce false positives.

Answer

Add a custom entity list for 'OrderID' patterns and configure the model to use it as a fallback.

Question 13

You are developing a solution that uses Azure AI Language custom named entity recognition (NER). You have labeled data with entities such as product names and quantities. After training, the model performs well on product names but poorly on quantities. You need to improve the model's performance on quantities without retraining from scratch. What should you do?

Accepted Answer

Add more labeled examples of quantities to the training dataset and retrain the model.. Custom NER improves with more relevant labeled data. Since quantities are underperforming, adding more labeled quantity examples and retraining is the direct way to boost the model's learning for that entity. Adjusting thresholds, switching to prebuilt models, or adding languages do not address the lack of training signal for quantities.

Answer

Enable multilingual support in the project and add examples in other languages.

Answer

Increase the model's confidence threshold for quantities in the project settings.

Answer

Switch to a prebuilt entity extraction model that includes quantity as a built-in entity.

Question 14

You are using Azure AI Language to extract key phrases from technical support tickets. The tickets are in English and contain many domain-specific terms such as 'Kubernetes', 'kubectl', and 'YAML'. You notice that the key phrase extraction is not identifying these terms. What should you do to improve the extraction?

Accepted Answer

Switch to using Azure AI Language custom named entity recognition (NER) to extract the domain-specific terms.. Key phrase extraction is a pre-trained feature that may not recognize domain-specific jargon. Custom NER in Azure AI Language lets you train a model with labeled examples to identify specialized terms like 'Kubernetes' and 'kubectl'. This directly improves extraction of the required technical terms. Other options do not provide a way to customize the extraction for niche vocabulary.

Answer

Increase the 'model-version' parameter in the key phrase extraction API call.

Answer

Use a custom text classification model to classify tickets by technology, then extract key phrases from each category.

Answer

Enable the 'logging' option in the Azure AI Language resource to capture unrecognized terms for retraining.

Question 15

You are building an Azure AI Language solution that must detect personally identifiable information (PII) in support tickets and redact it before storage. The tickets are written in a mix of English and German, and you must return the redacted text along with the categories of detected entities. You call the REST API endpoint for PII detection. Which request parameter should you include to ensure the service returns the redacted text in addition to the entity categories?

Accepted Answer

Include the parameter that requests the redacted text in the response, such as setting the appropriate response content option.. To obtain redacted text from the PII detection feature, you must explicitly request it in the API call. The service then returns a redacted version of the input text together with the entity categories and positions. Parameters that filter categories, select model versions, or define redaction style do not cause the redacted text to be included, so the request must enable the response content option that returns redacted output.

Answer

Set piiCategories to include all categories and set stringIndexType to TextElements_v8.

Answer

Set the model-version parameter to the latest generally available version.

Answer

Set the redactionPolicy parameter to a policy that specifies the redaction character or mask.