This document discusses deep distillation from text, which involves extracting more complex information than just entities and sentiment. It describes extracting expressions, entities, aspects, and contextual sentiment. The approach uses natural language processing, machine learning classifiers trained on engineered features, and leveraging existing knowledge sources and ontologies. Case studies show deep distillation applied to health informatics, retail surveys, and risk assessment from biomedical literature. The method aims to provide more actionable insights from unstructured text data.