Model training method, model application method and related device

The embodiment of the invention discloses a model training method, a model application method and a related device, because the feature value of a fixed value class feature is relatively fixed, even if the content information amount corresponding to a certain sample content is relatively small, the feature value of the fixed value class feature is relatively fixed. Each feature value of the fixed value class features also has more corresponding content information for model training, so that values of the non-fixed value class features are replaced by default values in the model training process, and comparison vectors are generated for comparison training; through the first loss function, the model learns how to generate the vector expression capable of reflecting the sample content specificity based on the fixed value class features, the expression of the fixed value class information on the content specificity during content sorting is enhanced, and the model can have relatively accurate content expression