6.3.3. Multi-Language Question Answering

💡 First Principle: Maintaining separate knowledge bases per language multiplies your content management burden—every FAQ update must be replicated across all languages. Multi-language QA solves this by accepting questions in any supported language and matching them against a single knowledge base (typically in English), then returning answers that can be translated on-the-fly. The exam tests whether you understand this is about operational efficiency (one KB to maintain), not just translation capability.

🔧 Implementation Reference: Multi-Language QA

Configuration Options:

Multi-Language Pattern:

from azure.ai.language.questionanswering import QuestionAnsweringClient

client = QuestionAnsweringClient(endpoint=endpoint, credential=AzureKeyCredential(key))

# Query with language hint
response = client.get_answers(
    question="¿Cuál es la política de devoluciones?",
    project_name="RetailFAQ",
    deployment_name="production",
    language="es"  # Spanish
)

Creating Multi-Language Knowledge Base:

# Import sources in multiple languages
sources = [
    {"url": "https://contoso.com/faq/en", "language": "en"},
    {"url": "https://contoso.com/faq/es", "language": "es"},
    {"url": "https://contoso.com/faq/fr", "language": "fr"}
]

⚠️ Exam Trap: For multi-language support, you can either enable "answer in same language" OR use the language parameter—not both simultaneously.

Selecting the Right Custom Language Approach

Use this decision tree to select between CLU and Custom Question Answering:

Pre-built vs. Custom Language Comparison