The project aims to develop a linguistically-aware supervised machine learning approach for cross-lingual political argument mining in user-generated text on Facebook, Youtube, Telegram, and other social media. The effort leverages the advances of argumentation theory in order to design a multi-dimensional argumentation scheme applicable to heterogeneous social media sources. The innovative part of the project consists in designing a new argument mining model for detection of political topics, aspects, stances, and sentiment. The ultimate goal for creating the model is to provide salient insights into public opinion and narratives on pressing political and social issues in multiple languages.