Model training method applied to text recognition and text recognition method and device

The invention provides a model training method applied to text recognition and a text recognition method and device, and relates to artificial intelligence, in particular to the fields of natural language processing, deep learning, semantic analysis and the like. According to the specific implementation scheme, a first to-be-trained text set comprising texts without category labels and a second to-be-trained text set comprising preset second category labels are obtained, wherein the second to-be-trained text set comprises a plurality of text subsets; training the initial model based on the first to-be-trained text set to obtain a first model; repeating the following steps, wherein the initial value of i is 1: inputting the ith text subset into the first model to obtain a trained first model; determining that the trained first model is a new first model, and determining that the value of i is i + 1; and determining that a new first model obtained when a preset condition is met is a text recognition model. The