About ===== Using CMUSphinx, an open-source speech recognition engine, we developed a fully voice-controlled Android application. Our app addresses the lack of spoken-language practice available in electronic form. Current applications, namely Duolingo, have a very weak speaking module that is both difficult and unengaging to use. Our application solves this problem by taking users through everyday (actually useful) conversation topics, including the ordering of a coffee. One issue we faced is the fact that originally you had to speak the phrase with a certain idiolect and speed, but we realized this was extremely difficult to achieve (and also a frustrating issue in the Duolingo world). To overcome this, we integrated a 'fuzzy strings' cosine similarities-based algorithm to allows users to pass a stage if they achieve a certain threshold of speaking accuracy. In the future we hope to implement a version with multiple languages, cross-platform usage, and a proliferation of interesting situations and conversations. We could also use modern-day machine learning algorithms to train an AI to a specific user's idiolect (i.e. speaking style) and better recognize unique phonetic pronounciations in a person-to-person context. You can demo this app by cloning to your own repository and running it on an Android device. This app is optimized for the Google Pixel 3.