| Abstract [eng] |
The mobile application market is enormous and most applications support multiple languages. However, supporting or implementing each new language requires significant financial and human resources. Every language must be thoroughly tested to ensure that the application meets the required standards and guarantees user satisfaction. Therefore, testing is extremely important in ensuring that everything is translated correctly into different languages and that no errors are made. Nevertheless, this process is usually performed manually, which is time-consuming, and human error often results in unnoticed defects. This thesis primarily focuses on the analysis of three technologies and their capabilities in testing various mobile application localization defects and proposing possible fixes for them. The work examines optical character recognition, object detection models, and multimodal artificial intelligence models. Before the experimental part, a survey will be conducted to identify the current situation and the problems faced by different software development teams. Finally, an experimental study of all the mentioned technologies will be carried out, evaluating their capabilities, limitations, and possible improvements that could lead to better results. Based on the obtained results, it will be possible to determine whether localization defect detection is a realistically achievable outcome. |
