Based on the strong classification feature recognition algorithm, the calculation algorithm of a text semantic similarity is studied with the performance evaluation in this paper. In order to achieve a general algorithm for this function, the semantic function library based on a semantic recognition code as a comparison object is designed. It drives the algorithm modules of two fuzzy neuron deep convolution machine learning, and between these two processes of machine learning, a rigid algorithm based on Fourier transform frequency domain feature is extracted. Finally, a more complex machine learning general algorithm is realized by the use of external data fuzzy algorithm and de-fuzzy algorithm before and after the algorithm module. It is also a technical innovation in this paper. Through the performance evaluation based on the subjective evaluation of volunteers, it is found that the system focuses on the text semantic similarity evaluation of the Chinese language, and achieves a comparison result of 81.78% of the artificial judgment accuracy rate, and only 5.52% of the volunteers believe that the system judgment result is completely different from that of manual judgment.