A semi-supervised novel recommendation algorithm

Traditional text processing models have been studied with short texts, but there are few studies for long texts recommendation. Novels as the long texts have higher preprocessing dimensions, more semantic textual relationships and complex relationship of characters compared with short texts, it makes long text recommendation difficult. In order to address the novel recommendation issue, this paper proposes a semi-supervised novel recommendation algorithm with tag-topic model. In the paper, we build a tag list set, finds sample data containing test data tag elements in the sample set, and performs topic model training on the text content of the sample data to obtain the topic distribution vector. Combine the sample set and test set data, topic model is used to perform training on its text content to obtain the topic distribution vector. The cosine similarity calculation is performed on the topic distribution vector, and the recommended data list is obtained through Top5 calculation. The experimental results show that tag-topic model recommendation method not only can obtain novel recommendation results by long texts recommendation, but also helps solve the time-consuming problem based on tag recommendation, and effectively recommends interesting novels for readers.