Abstract
A product is not useful unless it is usable. This statement can be universally applied to any products designed for human use including software systems. Unfortunately, we still see many common applications in our daily lives that are not so user-friendly. anyOCRWeb and anyTrain are web-based applications which offer users with many great features. Both are exceptional attempts strived for better user experience of anyOCR, a high-accuracy commandline interface Optical Character Recognition (OCR) system. Nevertheless, the effort must be continued. This thesis aimed to further improve the user experience of both systems by applying User-Centered Design approach. I focused primarily on OCR pipeline creation and visualization, as well as, output preview in anyOCRWeb, and clustering annotation for unsupervised OCR training in anyTrain. I began with the analysis of the context of use to grasp a good understanding of the very people who would be using the systems. I then specified the requirements, generated design solutions, and implemented them. Throughout the process, evaluation played an important role to ensure that this master’s thesis was moving in the right direction. Frequent usability testing was conducted, both formally and informally, regardless of the tested articles: sketches, prototypes, or real systems. Formal usability experiments were organized to quantify the improvement in usability. In a cross-over withinsubjects experiment, System Usability Scale (SUS) was chosen as a tool to measure perceived usability of the old and the new design. It was found that, on a scale of 0 to 100, there was a statistically significant rise in usability from 45.83 points in the old version to 75.14 points in the new version. In comparison with other web interface systems, the new user interface was also considered above average in terms of usability.