The invention discloses a multi-tag text classification method and system. The method comprises: determining to-be-checked tag subsets for a to-be-classified document; performing word segmentation processing on the to-be-classified document; obtaining a likelihood of a current calculated word in correspondence to each tag in a current check tag subset; performing linear weighting on the likelihood of the current calculated word in correspondence to each tag in the current check tag subset to obtain a weighted likelihood of the current calculated word in correspondence to the current check tag subset; determining each weighting coefficient for linear weighting, so that a continued product of weighted likelihoods of all words is maximal, and the maximum continued product serves as a likelihood of the to-be-classified document in correspondence to the current check tag subset; and in the to-be-checked tag subsets, selecting a tag subset with a maximum posteriori probability as a classification result of the to-be-classified document. Compared to an existing multi-tag classification method and system, the multi-tag classification method and system disclosed by the invention have the characteristics of high overall performance and small calculation amount.