Based on the description of the method for eliminating the ambiguity of the neural network polysemy: the semantic environment in which the ambiguous words are located is converted into a vector of words by intercepting a certain length in the neural network, and the iterative calculation is repeated by the neural network. Weights and offsets are adjusted to accurately classify the polysem, giving the most appropriate value.