AI-102 Implement natural language processing solutions Practice Test 3 — 15 Questions

Question 1

You are developing a custom text classification model using Azure AI Language. The model must classify customer support tickets into 15 categories. You have 10,000 labeled examples. After training, the model shows 95% accuracy on the test set but only 60% on a small sample of new tickets. What is the most likely cause?

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

Accepted Answer

The model is overfitting to the training data.. The model's 95% accuracy on the test set versus 60% on new tickets is a classic symptom of overfitting. In Azure AI Language custom text classification, overfitting occurs when the model learns noise and idiosyncrasies of the training data rather than generalizable patterns, causing poor performance on unseen data. The high accuracy on the test set but sharp drop on new tickets indicates the model memorized the training distribution and fails to generalize.

Answer

The model's confidence threshold is set too low.

Answer

The training data is not representative of the new tickets.

Answer

There is data leakage between the training and test sets.

Question 2

You need to extract key phrases from a large collection of customer reviews using Azure AI Language. The solution should be cost-effective and process up to 1,000 documents per day. Which pricing tier should you choose?

Accepted Answer

Standard (S) tier. The Standard (S) tier is the correct choice because it supports up to 1,000 documents per day for key phrase extraction, which matches the stated requirement. The Free (F0) tier is limited to 5,000 text records per month (roughly 167 per day), making it insufficient for 1,000 documents daily. The Premium (P) tier is designed for high-throughput scenarios (millions of documents per day) and is overkill for this volume, while the Basic (B) tier does not exist for Azure AI Language's key phrase extraction API.

Answer

Free (F0) tier

Answer

Premium (P) tier

Answer

Basic (B) tier

Question 3

A healthcare organization uses Azure AI Language to analyze clinical notes. They need to detect protected health information (PHI) such as patient names and dates of birth, and also identify medical conditions. Which Azure AI Language feature should they use?

Accepted Answer

Health Text Analytics (preview) in Azure AI Language. Health Text Analytics (preview) in Azure AI Language is the correct feature because it is specifically designed to extract and structure medical information from unstructured clinical text, including protected health information (PHI) such as patient names and dates of birth, as well as medical conditions. This preview API extends the Text Analytics for health capabilities with enhanced PHI detection and entity linking, making it the appropriate choice for healthcare compliance and clinical analysis.

Answer

Conversation Analysis

Answer

Custom Text Classification

Answer

Text Analytics for health (general availability)

Question 4

You are using Azure AI Language to analyze sentiment in customer feedback. The analysis returns a sentiment label of 'mixed' for a review that contains both positive and negative statements. The overall sentiment score is 0.75 (positive). What does this indicate?

Accepted Answer

The text contains both positive and negative sentiments.. The 'mixed' sentiment label indicates that the text contains both positive and negative statements. The overall sentiment score of 0.75 (positive) reflects the aggregate confidence that the text is positive, but the label 'mixed' is assigned when the model detects significant conflicting sentiments, meaning the text is not uniformly positive or negative. This is a key behavior of Azure AI Language's sentiment analysis, which provides both a label and a score for the overall document.

Answer

The text is negative overall.

Answer

The text is neutral overall.

Answer

The analysis is inconclusive.

Question 5

You are deploying a custom named entity recognition (NER) model using Azure AI Language. The model must extract product codes that follow a specific pattern (e.g., 'PRD-12345'). You have 5,000 labeled examples. After training, the model extractor works well on development data but fails to extract product codes from new data. What is the most likely issue?

Accepted Answer

The model is overfitting to the training data.. The model performs well on development data but fails on new data, which is the classic symptom of overfitting. In Azure AI Language custom NER, overfitting occurs when the model memorizes the training examples—including noise or specific patterns—rather than generalizing to the underlying product code pattern. With 5,000 labeled examples, the dataset size is likely sufficient, but the model may have learned spurious correlations that do not hold in unseen data.

Answer

The training data size is insufficient.

Answer

The product code pattern is too complex for the model to learn.

Answer

The labeling is inconsistent across the dataset.

Question 6

You need to analyze customer service call transcripts to identify common issues. The solution must extract key phrases, detect sentiment, and identify the language used. The transcripts are stored in Azure Blob Storage. Which Azure AI Language feature should you use to process them asynchronously?

Accepted Answer

Async Text Analytics API. The Async Text Analytics API (now part of Azure AI Language's asynchronous operations) is designed for batch processing of documents stored in Azure Blob Storage. It supports key phrase extraction, sentiment analysis, and language detection in a single asynchronous call, making it ideal for processing large volumes of customer service call transcripts without blocking the application.

Answer

Conversational Summarization API

Answer

Custom Text Classification API

Answer

Opinion Mining API

Question 7

You are building a custom question answering solution using Azure AI Language. Which TWO actions are required to deploy the solution?

Accepted Answer

Create a Custom Question Answering project in Azure AI Language.. Option A is correct because a Custom Question Answering solution in Azure AI Language must begin with creating a Custom Question Answering project, which holds the knowledge sources (question-answer pairs, URLs, and files) that define the solution. Option D is correct because after building and testing the project, you must deploy it to an Azure AI Language resource so the runtime endpoint becomes available for client applications to query. Options B, C, and E are not required: Custom Question Answering does not require training a custom model via the Azure AI Language training API, QnA Maker is the legacy predecessor service and is not needed, and publishing to a web app bot is an optional consumption choice rather than a deployment requirement.

Answer

Train a custom model using the Azure AI Language training API.

Answer

Create a QnA Maker service in the Azure portal.

Answer

Publish the project to a web app bot.

Question 8

You are designing a solution to detect personally identifiable information (PII) in documents using Azure AI Language. The solution must also handle documents in multiple languages. Which THREE features should you use?

Accepted Answer

The language detection feature of Azure AI Language.. Option A, the language detection feature of Azure AI Language, is correct because the solution must handle documents in multiple languages, and language detection identifies the language of each document so the correct language-specific model or processing path can be applied. Option C, the PII detection feature of Azure AI Language, is correct because it is the core capability that identifies and categorizes personally identifiable information such as names, addresses, phone numbers, and account numbers in text. Option E, the multilingual support in Azure AI Language PII detection, is correct because PII detection natively supports multiple languages, allowing the same PII model to detect entities across documents in different languages without building separate pipelines. Option B, the conversation summarization API, is not relevant because it summarizes conversations rather than detecting PII. Option D, Text Analytics for health, is not relevant because it extracts medical entities and relations from clinical text, not general PII across multilingual documents.

Answer

The conversation summarization API.

Answer

The Text Analytics for health feature.

Question 9

You are debugging a CLU application. The JSON above shows a request to the Azure AI Language runtime API. The response returns an intent of "BookFlight" with a confidence of 0.95, but no entities are extracted. The training data includes entities like "Location" and "DateTime". What is the most likely cause?

Accepted Answer

The request is missing the 'verbose' parameter set to true.. The CLU runtime API, by default, returns only the top intent and does not include extracted entities unless the 'verbose' query parameter is set to true. Without this parameter, the response omits the entities array even if the model was trained to extract them, which explains why the intent is returned with high confidence but no entities appear.

Answer

The deployment name is incorrect.

Answer

The model was not trained on entity extraction.

Answer

The utterance does not contain any entities.

Question 10

You deploy the ARM template above to create an Azure AI Language resource. After deployment, you try to use the custom question answering feature but it is not available. What is the most likely reason?

Accepted Answer

The 'kind' property is set to 'TextAnalytics', which does not enable custom question answering.. The 'kind' property in the ARM template is set to 'TextAnalytics', which provisions a general-purpose Text Analytics resource. Custom question answering is a feature of the Azure AI Language service that requires the resource to be created with the 'kind' set to 'Language' or 'ConversationalLanguageUnderstanding' (depending on the API version). Because the resource is of the wrong kind, the custom question answering capability is not available, even though the deployment succeeds.

Answer

The apiVersion is too old to support custom question answering.

Answer

The SKU is set to S, but custom question answering requires a higher SKU.

Answer

The location does not support custom question answering.

Question 11

You are building a conversational AI solution that must handle multiple intents in a single user utterance. Which Azure AI feature should you use?

Accepted Answer

Use the orchestration workflow feature in Conversational Language Understanding.. The orchestration workflow feature in Conversational Language Understanding (CLU) is designed to handle multiple intents within a single user utterance by connecting to different projects (e.g., CLU, QnA Maker, or custom question answering) and routing the utterance to the appropriate service. This allows the solution to parse complex utterances that may trigger multiple intents across different domains, which is exactly what the scenario requires.

Answer

Use QnA Maker with multiple QnA pairs.

Answer

Use a single Conversational Language Understanding project with multiple intents.

Answer

Use Azure Bot Service with multiple dialogs.

Question 12

Your NLP solution uses custom text classification in Azure AI Language. You need to improve the model's accuracy. Which action should you take?

Accepted Answer

Add more labeled training data with balanced classes.. Custom text classification in Azure AI Language relies on supervised learning, where model accuracy is directly proportional to the quantity and quality of labeled training data. Adding more labeled examples with balanced class distributions reduces bias, improves generalization, and helps the model learn distinguishing features more effectively, which is the most impactful action for improving accuracy.

Answer

Increase the number of training epochs.

Answer

Decrease the batch size.

Answer

Use a different pretrained model.

Question 13

Which TWO actions can you take to improve the performance of a Conversational Language Understanding model?

Accepted Answer

Add more varied utterances to each intent.. Option A is correct because adding more varied utterances to each intent gives the model broader lexical and syntactic coverage, which improves its ability to generalize and correctly classify user input during training and prediction. Option C is correct because the Evaluate feature in Conversational Language Understanding reports metrics such as precision, recall, and F1 score per intent and highlights misclassified utterances, letting you identify weak intents and add or fix training data to raise model accuracy. Option B is not appropriate because simply reducing the number of intents does not inherently improve model quality and may remove required functionality. Option D is incorrect because changing the Azure region affects latency and data residency, not the model's trained performance. Option E is incorrect because disabling active learning removes the ability to surface low-confidence utterances for labeling, which typically degrades rather than improves model performance over time.

Answer

Reduce the number of intents.

Answer

Change the Azure region of the resource.

Answer

Disable active learning.

Question 14

Which THREE factors should you consider when selecting a region for an Azure AI Language resource?

Accepted Answer

Data residency requirements.. Option C (Data residency requirements) is correct because Azure AI Language processes text data in the region where the resource is deployed, so compliance, sovereignty, and regulatory obligations may dictate that data stays within a specific geography. Option D (Language and feature availability) is correct because not every Azure AI Language feature (for example, certain custom question answering, summarization, or entity extraction capabilities) and not every supported language is offered in every region, so you must verify the specific feature/language pair exists in the target region. Option E (Latency to the application's users) is correct because choosing a region geographically close to the calling application or end users reduces network round-trip time for synchronous REST/SDK calls to the Language service endpoint. Option A (Pricing differences) is not a valid selection factor here because Azure AI Language pricing is based on the SKU/transaction model and is generally consistent across regions rather than being a region-selection criterion. Option B (Availability of the Free tier SKU) is not a valid selection factor because the Free (F0) tier is intended for evaluation and is not the basis for production region selection; capacity and feature support matter more.

Answer

Pricing differences.

Answer

Availability of the Free tier SKU.

Question 15

You submit the request shown in the exhibit to the Azure AI Language service. What will the response contain?

Accepted Answer

Both key phrases and sentiment scores.. The exhibit shows a request to the Azure AI Language service with the `showStats` parameter set to `true` and the `opinionMining` parameter set to `true`. When `opinionMining` is enabled, the service returns both key phrases and sentiment scores (including aspect-based sentiment) for the analyzed text. Therefore, the response contains both key phrases and sentiment scores, making option C correct.

Answer

Only extracted key phrases.

Answer

Entity recognition results and sentiment scores.

Answer

Only sentiment scores.