Research on Text Summarization Generation and Classification Method Based on Deep Learning

The information extraction and classification of text is of great significance in many aspects such as text retrieval, news statistics, and literary works analysis. However, the exponential increase of texts generated in the information age poses a huge challenge to the sort texts. Based on deep learning technology, this paper generates abstracts for long texts, and then classifies texts based on the content of the abstracts. Since the abstract contained main information, the primary information of the document is extracted and stored in the abstract through automatic summarization, and then the classification may have a better effect. This paper adopts the improved model Pointer-generator network based on the Sequence-to-sequence attentional model to generate generative summaries, and then builds a TextCNN model to classify the generated summaries at the character level. Those not only generate the abstract of the document but also classify the document. The fusion of the two can more effectively reduce the search pressure. The experiment shows that the proposed method has achieved the desired results and could be utilized in many aspects in the future.