The embodiments of the application relate to the field of artificial intelligence, in particular, to a training method and apparatus for a speech recognition model. The training method comprises: a speech is inputted into a speech recognition model; N texts corresponding to the speech are obtained at the output side of the speech recognition model; each of the N texts matches a plurality of textsstored in a preset database to obtain N matching degree values corresponding to the N texts; according to the N matching degree values and a preset condition, one text corresponding to the matching degree value meeting the preset condition among the N matching degree values is determined to be target text; and the speech recognition model is trained by using the speech and the target text as training data of the speech recognition model. Therefore, deletion is carried out directly by the N texts obtained by the speech recognition model and thus the target text is determined, so that the targettext can be obtained without manual marking and thus the labor cost is saved.