Write a very high-quality and detailed summary of the paper that describes the paper in a way that a human can fully understand. The summary should cover the problem, the proposed solution and highlight the main contributions of the paper.
Here is a detailed summary of the paper:
Problem:
- Public code review (PCR) is an important software engineering practice where developers submit code for review by the public community.
- Two key tasks that impact PCR quality are 1) request necessity prediction - predicting whether a review request is necessary, and 2) tag recommendation - recommending relevant tags for the request.
- Existing methods require designing specialized models for each task, which is time-consuming.
Proposed Solution:
- The paper proposes a unified framework called UniPCR to jointly address both tasks using prompt tuning.
- It uses text prompt tuning to reformulate the tasks as masking language model (MLM) generation by designing descriptive templates.
- It uses code prefix tuning to optimize a small vector prefix for code representation.
- Both prompt tuning strategies are combined under a unified framework to predict necessity and recommend tags.
Key Contributions:
- Novel unified framework UniPCR to jointly train request necessity prediction and tag recommendation for PCR via prompt tuning.
- Reformulates tasks as generative MLM problems instead of specialized discriminative models.
- Empirically demonstrates UniPCR outperforms state-of-the-art by 8-28% on necessity prediction and tag recommendation over strong baselines.
- Ablation studies validate benefits of unified text and code prompt tuning strategies.
- Showcases potential of using single unified framework for multiple PCR quality assurance tasks.
In summary, the paper proposes a novel unified PCR framework UniPCR using prompt tuning to jointly train necessity prediction and tag recommendation in a generative setup. Experiments demonstrate clear improvements over specialized discriminative methods.