A method for categorizing documents may include receiving topic information for a source document, the information including at least one topic and a weight for each topic, where the topic relates to content of the source document, and the weight represents how strongly the topic is associated with the source document. Similarity scores may be generated based on a weight of each topic in the source document and the weight of the same topic in each document within one or more sets of documents, where each document in the one or more sets of documents comprises topic information. A confidence score may be generated, based on the similarity scores, for each of the document sets. One or more document sets may be selected based on the confidence scores and may be output to a user.