Develop an Android app that acts as your personal language trainer. Key requirements are: (1) The application is 100% speech based, i.e. all outputs are audios and all inputs are captured from the microphone; (2) the focus is 90% on speaking, i.e. grammar is sometimes corrected, but mostly it is about extending vocabulary and creating fluency; (3) the system remembers your state, learns from feedback and tries to understand the user’s needs as good as possible; (4) everything should run locally and be based on free models & libraries.
Hence, key hurdles are: (a) getting speech recognition work reliably (sound rendering & providing different speech personae should be no problem), (b) selecting an SLM that is able to understand, converse and teach (in at least some of the major languages of the world), but still runs on a contemporary smartphone, (c) minimizing battery consumption, (d) fine-tuning the system so that, in particular, the first sessions after installation go well (talking about politics or rather the weather?). Furthermore, the app should be based on an extendible framework so that later, new features could be added (e.g. context-related chats based on the current location, etc.).
Requires good knowledge in app programming and an interest in language learning. If performed as a diploma thesis, must include a user-based evaluation. AI-based code generation is allowed for this task, as it will still be hard enough to get the content/interaction problem right.