Multiview formulation of data is required to construct coherent and contextualized user representation on social media. Our paper suggests a joint embedding model by incorporating users' social and textual information to learn contextualized user representation to understand users' lifestyle choices. In this work, we use well-being related tweets from Twitter, focusing on \textit{`Yoga' and `Keto diet'}. We compare our model's performance against several other modeling choices and our model outperforms those baselines. We illustrate data collection and annotation details as well as in-depth data analysis of users from different classes.