A method and device for correcting a search query term: when a query term in a search request is received by a search query term correction device (S201), first acquiring a correction candidate sub-term list for sub-terms in the query term (S202); then online rating in real-time, on the basis of a pre-established N-Gram language model, correction candidate terms spliced from the correction candidate sub-term list, acquiring a correction target term list (S203); and then finally outputting a final correction result on the basis of ratings of correction target terms in the correction target term list (S204). The search query term entered by a user is corrected in real-time on the basis of earlier analytic computations, the need for offline big data processing and a past correction database is obviated, reliance on large-scale search logs and user feedbacks is removed, correction capabilities are effectively increased for application scenarios such as a mobile application market.
