The invention provides an automatic abstract generation method and device based on a generative adversarial network, and the method comprises the steps: firstly obtaining question data from a questionand answer community, cleaning the data through a script, dividing a training set and a test set, building a dictionary, and generating a training set sentence vector through the dictionary; trainingsentence vectors in the training set by using a generative adversarial network model to obtain appropriate model parameters; and finally, performing automatic abstract generation of similar questionanswers on the test set by using the trained model. The technical effects of obtaining the data set in an unsupervised learning mode and improving the generation effect of the automatic abstract are achieved.