This paper studies the accuracy correction method of literary translation based on semi supervised learning, so as to improve the accuracy of Literary translation and reduce the translation error rate. Based on the word vector of recurrent neural network, the data preprocessing and feature extraction of translation of Literary works are realized by constructing a word alignment and segmentation model. Based on TF − IDF algorithm, the translation grammatical features of Literary works are extracted, and K-means clustering algorithm is used to detect the accuracy features. Based on semi supervised learning, mistranslation features are identified, and translation accuracy correction of Literary works is realized through semantic feature analysis. The results show that this method can detect the mistranslation features in the grammar feature sample set. The number of mistranslation features detected is almost the same as the actual number of mistranslation categories in the corresponding data set, and the comprehensive detection performance is high; We can distinguish the mistranslated grammar from the correct grammar through grammar mistranslation correction, and the overall correction accuracy is higher than 98% .