Argumentation mining and stance classification were recently introduced as interesting tasks in text mining. What is particularly interesting about these task is new challenges and consequently opportunities that they presents. In this paper, a new framework for argument tagging based on topic modeling, mainly Non-Negative Matrix Factorization, is proposed. These extracted arguments may subsequently be exploited for stance classification. Experiments on our collected corpus of news comments demonstrate the benefits of using topicmodeling for argument tagging. Furthermore, the statistical model that leverage automatically extracted arguments as features shows promising results.