Bash code annotation generation method based on semantic information enhancement

The invention provides a Bash code annotation generation method based on semantic information enhancement, and belongs to the field of computers. The problems that in a Bash annotation generation task, a data set is small, and all feature information output by CodeBERT cannot be fully utilized are solved. According to the technical scheme, the method comprises the following steps that S1, adversarial training is used in an embedded layer generated in a corpus, and adversarial samples are generated to enhance semantic information; s2, performing fine adjustment on CodeBERT by using word embedding and confrontation samples generated by an original corpus, and outputting 12 layers of representation information; s3, aggregating 12 layers of characterization information by using a Bi-LSTM + attention mechanism to generate fusion information; and S4, decoding by using a decoder and outputting the Bash annotation. The method has the beneficial effects that the model provided by the invention can capture and utilize