AI-102 Practice Question: Implement natural language processing solutions
Your company runs a global e-commerce platform. You are building a chatbot using Azure AI Language's conversational language understanding (CLU) to handle customer requests in multiple languages. The bot must support English, German, and Japanese. You have labeled training data in English only. The deadline is tight, and you want to minimize manual labeling. You also need to ensure that the bot can gracefully handle unsupported languages (e.g., French) by directing the user to a human agent. You have access to Azure AI Translator. Which approach should you take?
⚠ Common exam trap
This exam often tests the misconception that you must translate training data or build separate projects for each language, when in fact the multilingual option in CLU enables a single project to handle multiple languages with English-only training data, and the trap is that candidates overlook this built-in capability and choose more labor-intensive options like translation or separate projects.
Answer choices
Why each option matters
Answer the question above first, then reveal the full breakdown to understand why each option is right or wrong.
Correct answer & explanation
✓
Use a single CLU project with multilingual option enabled, train on English data only. Configure the bot to detect the language of user input; if it is English, German, or Japanese, route to CLU; otherwise, route to a human agent.
Azure AI Language's CLU supports a multilingual option that allows a single project to handle multiple languages without requiring translated training data. By enabling this option and training on English data only, the model can generalize to German and Japanese due to shared multilingual embeddings. The bot can then detect the user's language and route unsupported languages like French to a human agent, minimizing manual labeling while meeting the deadline.
Answer analysis
Option-by-option breakdown
For each option: why learners choose it and why it is or isn't the right answer here.
- ✗
Use a single CLU project with English data only. Translate the English training data into German and Japanese using Azure AI Translator, then train a single multilingual model by including the translated data.
Why it's wrong here
Translating English training data into German and Japanese via Azure AI Translator introduces translation artefacts that degrade the CLU model's ability to learn natural language patterns in those languages, as the translated utterances lack native idiomatic phrasing and cultural context. This approach is tempting because it appears to reuse existing English labels without manual effort, and it would be correct if the target languages were structurally similar to English and the chatbot only needed to match keywords rather than understand nuanced intents across languages.
- ✗
Use a single CLU project with English data only. Before calling CLU, translate non-English user input to English using Azure AI Translator. For unsupported languages, detect language and route to human agent.
Why it's wrong here
Translating user input can introduce errors and latency; the multilingual model can understand multiple languages without translation.
- ✓
Use a single CLU project with multilingual option enabled, train on English data only. Configure the bot to detect the language of user input; if it is English, German, or Japanese, route to CLU; otherwise, route to a human agent.
Why this is correct
The multilingual option allows the model to predict intents in English, German, and Japanese without additional labeled data. Language detection ensures unsupported languages are handled appropriately.
- ✗
Build separate CLU projects for English, German, and Japanese. Label training data in each language by translating the English data using Azure AI Translator.
Why it's wrong here
This requires labeling effort in each language, even with translation, and maintaining multiple projects. It does not minimize manual labeling.
Go deeper
Related to this question
About these practice questions
One of 945 original AI-102 practice questions on Courseiva, each with a full explanation and wrong-answer analysis — not exam dumps or protected exam content. Learn why practice questions differ from exam dumps →
JA
Written by Johnson Ajibi, MSc IT Security
Senior Network & Security Engineer · founder of Courseiva
This AI-102 practice question is part of Courseiva's free Microsoft certification practice question bank. Courseiva provides original exam-style practice questions with explanations, topic-based practice, mock exams, readiness tracking, and study analytics to help learners prepare for the AI-102 exam.