This document discusses how grounded theory can be applied to analyze large datasets from social media platforms like Twitter. It proposes combining both qualitative and computational methods like machine learning. Specifically, it suggests using machine learning like latent Dirichlet allocation to identify topic clusters in Twitter data, which can then inform the qualitative coding categories used to analyze content, profiles and other metadata. The document advocates an emergent approach to coding to build conceptual knowledge from Twitter data, and emphasizes the importance of being reflexive and considering multiple perspectives.