Based on the characteristics of digital literature, this paper proposes a new trend in the development of autonomy and human-computer interaction in digital literary creation. In view of the changes in the path of literary dissemination, it describes the mechanism of interaction between the digitization of literature and the audience, as well as the content. Taking the integration of literary creation and technology as an entry point, the Skip-Gram model is used to realize the parameter training of Chinese vocabulary, and combined with the binary language model, the joint distribution probability of each group of words is found and the optimal word splitting scheme with the highest probability is obtained. The attention mechanism of the improved channel effectively enhances the performance of the network model for the sequence learning task and predicts the currently occurring words according to the contextual features, realizing the automatic generation of literary creation content. To assess the effectiveness of the new media technology application, quality assessment indicators are introduced. Literary creation content on the two platforms has a mean value of 3.4 and 3.425, which are both of good quality. The overall user activity was above 20 and reached a peak of 45.265 in August, with a year-on-year increase ranging from 0.648 to 1.589, a significant increase in activity, and the integration of new media technology significantly enhanced the enthusiasm for literary creation.