Classification model training method and device

The embodiment of the invention provides a classification model training method and device. The method comprises: acquiring a first sample set of a first field, wherein the first sample set comprisesmultiple first training samples, and each first training sample comprises sample input and a category label of a corresponding first classification task; obtaining a second sample set of a second domain, wherein the second sample set comprising a plurality of second training samples, each second training sample comprising a sample input and a category label of a corresponding second classificationtask, and the category labels of the two classification tasks have a mapping relationship; adding a plurality of second training samples in the second sample set into the first sample set, and converting category labels of second classification tasks included in the second training samples into category labels of the first classification tasks according to the mapping relationship to obtain a third sample set; and inputti