This paper proposes Knowledge-Guided Prompt Learning (KP-PCR), a novel method for predicting the need for developers' code review requests and recommending appropriate tags, to improve the open code review (PCR) system developed in the Software Question Answering (SQA) community. Unlike previous PCR research, which primarily focuses on the reviewer's perspective, KP-PCR focuses on improving developer request satisfaction. To achieve this, it performs two sub-tasks: request need prediction and tag recommendation. This sub-task involves tuning text prompts based on a Masked Language Model (MLM), and fine-tuning knowledge and code prefixes using a large-scale language model and program dependency graphs. The final results are output by the Answer Engineering module. Experimental results using the PCR dataset from 2011 to 2023 demonstrate that KP-PCR outperforms existing methods by 2.3% to 8.4% in request need prediction and 1.4% to 6.9% in tag recommendation. The code is available on GitHub.
