A Field-test of a Web Application for L2 Speaking Practice under an Uncontrolled Learning Environment for a Whole Semester

In second/foreign language speaking practice, there is limited time for students to practice, getting feedback from an experienced teacher in class. In order to remove the time constraint, computer assisted language learning systems are the first candidate, but few systems estimate learner's state, which includes their mental states and learning environments. For estimating it, it is natural to use specific peripherals or applications, but this leads to another constraint in learning environment. In addition, it is challenging to develop such a system because speaking practice, other than practice for other skills, essentially requires non-textual data. The distant goal of this study is to propose a system that generates feedback based on learners' states, without any constraints on time and place. With this goal, the purpose of this paper is to identify the basic requirements of a system for speaking under uncontrolled learning environment by a long-term field test. In this field test, a prototype system we developed was used for 15 weeks (or 1 semester) as online assignments by 28 international students taking a Japanese language course at a university. Our prototype was a browser-based web application, and learners practiced speaking using their own devices, e.g., PCs and smartphones. This field test was conducted under uncontrolled learning environment of each learner participating in the long-term online course, without ever meeting face-to-face. These settings significantly differ from those of previous studies, which are well controlled under some specific environment. Our prototype has worked with many devices, and enabled students, including those who was not able to enter our country due to the coronavirus disease 2019 pandemic, to practice speaking. Through the field test, we identified some problems, which can be happened for similar systems of speaking, and summarized into three categories: operation devices, voice recording, and data transmission. We propose a set of methods to solve these problems, which are useful for developing similar systems in actual learning environments.