Assessing Perceived Sentiment in Pull Requests with Emoji: Evidence from Tools and Developer Eye Movements
Assessing Perceived Sentiment in Pull Requests with Emoji: Evidence from Tools and Developer Eye Movements
复制标题
DOI：
10.1109/semotion52567.2021.00009
复制
发表时间：
2021-05
期刊：
影响因子：
--
通讯作者：
Kang-il Park;Bonita Sharif
中科院分区：
文献类型：
--
作者：
Kang-il Park;Bonita Sharif
该论文介绍了一项有关了解开发人员在Google Chrome中的各种元素上收集的，了解了二十四个Github拉动请求的开发人员如何阅读和评估情感。开发人员的任务是确定感知的情绪。分析工具的性能最高，另一方面，其预测的55.56％数据显示了开发人员最多的三个领域是评论主体，添加的代码和用户名（写评论的人）也表现出高度关注。与其余的评论文本相比，要在拉的请求评论主体中表情符号。
The paper presents an eye tracking pilot study on understanding how developers read and assess sentiment in twenty-four GitHub pull requests containing emoji randomly selected from five different open source applications. Gaze data was collected on various elements of the pull request page in Google Chrome while the developers were tasked with determining perceived sentiment. The developer perceived sentiment was compared with sentiment output from five state-of-the-art sentiment analysis tools. SentiStrength-SE had the highest performance, with 55.56% of its predictions being agreed upon by study participants. On the other hand, Stanford CoreNLP fared the worst, with only 5.56% of its predictions matching that of the participants’. Gaze data shows the top three areas that developers looked at the most were the comment body, added lines of code, and username (the person writing the comment). The results also show high attention given to emoji in the pull request comment body compared to the rest of the comment text. These results can help provide additional guidelines on the pull request review process.