You are building an NLP system that receives user messages in multiple languages, sometimes within the same thread. Some messages are short, informal, and code-switched, while others use non-Latin scripts or accented characters. You need a practical approach for handling multilingual text without breaking downstream classification quality.
How do you handle multi-lingual text in an NLP system?