Building and customizing solutions using Microsoft 365 Copilot APIs and tools
To create a multilingual agent in Copilot Studio that can recognize both Traditional Chinese (zh-TW) and Simplified Chinese (zh-CN), you need to ensure that your agent is set up to handle these specific language distinctions. While the agent can detect if the input is Chinese, distinguishing between Traditional and Simplified Chinese requires specific configuration.
- Language Detection: When setting up your language detection topic, ensure that your conditions are explicitly checking for the structured output that differentiates between Traditional and Simplified Chinese. You may need to verify that your prompt correctly identifies these languages in the output.
- Custom Variables: Instead of using a general variable like
Topix.x.structedOutput.language, you should ensure that your prompt's model response is configured to return specific identifiers for both Traditional and Simplified Chinese. This might involve adjusting the prompt instructions to clearly differentiate between the two. - Agent Settings: In your agent's settings, make sure that both Traditional and Simplified Chinese are added as separate languages. This will help the agent recognize and respond appropriately based on the detected language.
- Testing: After configuring the conditions and ensuring the correct language variables are in place, test your agent with inputs in both Traditional and Simplified Chinese to confirm that it can distinguish between them accurately.
If you continue to face issues, consider reviewing the documentation on setting up multilingual agents and language detection to ensure all configurations are correctly applied.
References: