Commonsense reasoning refers to the ability of evaluating a social situation\nand acting accordingly. Identification of the implicit causes and effects of a\nsocial context is the driving capability which can enable machines to perform\ncommonsense reasoning. The dynamic world of social interactions requires\ncontext-dependent on-demand systems to infer such underlying information.\nHowever, current approaches in this realm lack the ability to perform\ncommonsense reasoning upon facing an unseen situation, mostly due to\nincapability of identifying a diverse range of implicit social relations. Hence\nthey fail to estimate the correct reasoning path. In this paper, we present\nConditional SEQ2SEQ-based Mixture model (COSMO), which provides us with the\ncapabilities of dynamic and diverse content generation. We use COSMO to\ngenerate context-dependent clauses, which form a dynamic Knowledge Graph (KG)\non-the-fly for commonsense reasoning. To show the adaptability of our model to\ncontext-dependant knowledge generation, we address the task of zero-shot\ncommonsense question answering. The empirical results indicate an improvement\nof up to +5.2% over the state-of-the-art models.\n