I need assistance with splitting concatenated English,spanish and italian words using a Hugging Face pipeline. The goal is to get splitted words from domain name in the fastest time with inference on local machine. Training can be done on gpu machine. I need the code to obtain this result, not the result or the model itself. I need a well structured code with comments and documentation. I have a test dataset with a list of words to check results. Requirements: - Experience with Hugging Face models - Familiarity with text processing and NLP - Ability to produce required code with time (<0.5 secs on cpu) and quality (>90% on test dataset) constraints. Please share relevant experience and approach.