A method for information recommendation is described. Processing circuitry obtains a first text content associated with which a recommendation is to be provided, segments the first text content into multiple words, and predicts a distribution of the first text content in a multi-dimensional topic space based on the multiple words. Further, the processing circuitry calculates a correlation between the first text content and a second text content in an information pool according to the distribution of the first text content in the multi-dimensional topic space. Then, the processing circuitry determines whether to output the second text content as the recommendation associated with the first text content according to the correlation.