hi thanks for your interest in my article! Yes, it is generally a good idea to preprocess the text before you feed into any model for prediction. Depending on your specific application, you can do different preprocessing.

As for French text classification, you'll need a pretrained French model first. For example, this BERT model is pretrained on many languages including French for sentiment analysis: https://huggingface.co/nlptown/bert-base-multilingual-uncased-sentiment