We introduce a proposal to theoretically characterize Information Retrieval (IR) supporting metadata. The proposed model has its foundation in a classical approach to IR, namely vector models. These models are simple and implementations are fast, their term-weighting approach improve retrieval performance, allow partial matching, and support document ranking. The proposed characterization includes document and query representations, support for typical IR-related activities like stemming, stoplist application or dictionary transformations, and a framework for similarity calculation and document ranking. The classical vector model is integrated as a particular case in the new proposal.