An example dataset of interaction logs of software company employees

This is a sample piece from a dataset of interaction logs recorded from software company employees. The data set is recorded at an overseas branch office of a Japanese software development company. The subjects are 18 employees of this branch office, who have varying qualifications and responsibilities. The data collection campaign is carried out with the consent of the company. The subjects are informed in a clear manner about the nature and method of the research, and agreed to participate in the experiments. <br> The data set is composed of the employees' interactions logs, which are basically a registry of their interaction with the GUI. The recording software is a user activity monitoring tool called TaskPit [1], which is designed to be deployed particularly at software development companies. Specifically, it registers a log file containing the name of the active application (i.e. exe name), start and end time of its deployment, its window title, and number of left clicks, right clicks, and key strokes on the active window. We consider each line of the log file to arise from an action of the subject. Moreover, each action is considered to be associated with a single task of the subject (e.g. Programming) and the tasks associated with those actions are exactly what our study intends to unveil. A coder, who is a senior student at the department of computer science of Okayama University, carried out manual annotation by assigning a single task to each action (i.e. each line of the log file). To that end, he evaluated the information contained in the columns of the log file and selected one task from the set of potential tasks. Here, by taking into account the background information of the subjects (i.e. being employees of software development company) and the expectations and requirements of our corporate partner, the set of potential tasks is tailored to be comprised of Programming, Test, Documentation, Administration, and Leisure. Note that by documentation we refer to reading, writing or editing of project documentation. The data set is used to test the task estimation method proposed in our article submitted to Empirical Software Engineering (currently under review) [2]. The codes generated during the current study are publicly available at our repository [3]. References: [1] Suthipornopas P, Leelaprute P, Monden A, Uwano H, Kamei Y, Ubayashi N, Araki K, Yamada K, Matsumoto K (2017) Industry application of software development task measurement system: Taskpit. IEICE Transactions on Information and Systems (3):462–472 [2] Pellegrin F, Yücel Z, Monden A., Leelaprute P. , Estimating tasks of software company employees based on computer interaction logs, Empirical Software Engineering (under review) [3] Yücel Z, Software applications and custom codes. https://github.com/yucelzeynep/Task-estimation-from-activity-logs, 2020本数据集取自某日本软件开发公司海外分支机构的员工交互日志样本集。研究对象为该分支机构的18名员工，其资质与岗位职责各不相同。本次数据采集工作已获得公司许可，研究人员已向受试者清晰说明研究的性质与方法，受试者均同意参与本实验。 该数据集由员工的交互日志构成，本质上是其与图形用户界面（Graphical User Interface，GUI）交互的记录。数据采集使用名为TaskPit的用户活动监控工具[1]，该工具专为软件开发公司部署设计。具体而言，其生成的日志文件会记录活跃应用程序名称（即可执行文件名）、应用启动与结束时间、窗口标题，以及活跃窗口内的左键点击、右键点击次数与按键输入次数。我们将日志文件的每一行视为受试者的一次操作，且每一次操作均对应受试者的一项任务（例如编程），而这些操作关联的任务正是本研究旨在揭示的核心内容。 冈山大学计算机科学系的一名高年级学生作为编码员，对每一次操作（即日志文件的每一行）手动标注对应的单一任务。为此，他会结合日志文件各列包含的信息，从候选任务集中选取一项。结合受试者的背景（即软件开发公司员工）与合作企业的期望与要求，候选任务集限定为编程、测试、文档编写、行政工作与休闲活动。需说明的是，本文中“文档编写”涵盖项目文档的阅读、撰写与编辑工作。 本数据集用于验证我们提交至《实证软件工程》（目前处于审稿阶段）的论文中提出的任务估算方法[2]。本研究中生成的代码已公开至我们的代码仓库[3]。 参考文献： [1] Suthipornopas P, Leelaprute P, Monden A, Uwano H, Kamei Y, Ubayashi N, Araki K, Yamada K, Matsumoto K (2017) Industry application of software development task measurement system: Taskpit. IEICE Transactions on Information and Systems (3):462–472 [2] Pellegrin F, Yücel Z, Monden A., Leelaprute P. , Estimating tasks of software company employees based on computer interaction logs, Empirical Software Engineering (under review) [3] Yücel Z, Software applications and custom codes. https://github.com/yucelzeynep/Task-estimation-from-activity-logs, 2020