Monte Carlo Tree Search-Driven Photovoltaic Text Rewriting

Rewriting photovoltaic (PV) technical texts is essential for improving clarity, accuracy, and adherence to industry standards. However, automated PV text rewriting remains challenging due to the need for domain-specific precision, logical coherence, and semantic preservation. This paper introduces a Monte Carlo Tree Search (MCTS)-driven framework for PV text rewriting, integrating structured prompting with iterative refinement. Our approach first employs a pretrained language model guided by fine-grained strategy prompts to generate domain-adapted draft texts. Then, MCTS systematically explores and refines text variations, optimizing multiple quality dimensions, including technical accuracy, terminology consistency, fluency, and readability. By dynamically adjusting scoring weights, our method ensures continuous improvement in text quality. Experimental results demonstrate that the proposed framework significantly improves PV text rewriting compared to conventional methods, offering a scalable and adaptive solution for professional technical documentation.