Although the existing research on educational robots has exhibited the assistance for EFL learners' English skills, the evidence which shows robot-assisted systems' effect on adult learners' English read-aloud is still rare. Nevertheless, read-aloud is still treated as a useful approach in English classes for speech pronunciations in particular in Asia. This study aims at the design of a system that uses a robot as a tutor equips automatic speech recognition to diagnose learners' errors while they are reading aloud the English passages. The learner is able to practice and improve pronunciations via the diagnosis. An experiment designed with a pretest, a posttest, and a delayed posttest was conducted to evaluate the proposed robot-tutoring system. 19 university students in Taiwan enrolled in the experiment, and learned with the system for 90 minutes per round, with a total of 2 rounds of self-learning. The results showed the participants' accuracy of read-aloud in the delayed posttest was significantly better than those of the pretest and immediate posttest. In addition, when looking into the investigation of the participants' perceptions of the system, most were impressed by the system, and also agreed the system was a useful and helpful tool to help reading English passages aloud. Therefore, this study provides evidence of the effects of robot-tutoring approach on adults' English read-aloud in Taiwan.