Figure8

Figure 8. Repetitive self-supervised learning process. (a) is the process of learning a small number of input instructions that conform to a variety of grammatical structures. Blue, green, and red correspond to three different grammatical structures, and dark samples are entered before light samples. (b) is the process of learning a large number of instructions that conform to a grammatical structure. The light color line is to track the learning process of five instructions in random sampling, and the dark color line is to average the number of iterations needed to extract the named entity from all samples in one cycle. The horizontal axis of both graphs represents the number of times a sample appears in the model, and the vertical axis represents the number of SSL iterations needed to accurately implement the NER of input instructions.