Optimal way to chunk word document for RAG(semantic chunking giving bad results)

Is there any code for CLAREDI, that I can refer to ?