A system and method for providing default hierarchical training for social indexing is provided. Articles of digital information for social indexing are maintained. A hierarchically-structured tree of topics is specified. Each topic includes a label that includes one or more words. Constraints inherent in the literal structure of the topic tree are identified. For each topic in the topic tree, a topic model that includes at least one term derived from the words in at least one of the labels is created. The topic models for the topic tree are evaluated against the constraints. Those of the topic models, which best satisfy the constraints are identified.