Mongolian Automatic Text Summarization Method Based on Pre-trained Model and Improved TextRank

At present, there is limited research on automatic Mongolian text summarization, especially using mainstream methods.The existing TextRank algorithm only considers the similarity between sentences, ignoring the characteristics of the sentence itself.In this paper, a Mongolian automatic text summarization method called IMNUBERT-mnTextRank, based on a pre-trained model and an enhanced Tex-tRank algorithm, is proposed.The information from the Mongolian external knowledge base is incorporated into the TextRank algorithm in the form of sentence vectors to enhance the accuracy of similarity calculations between sentences.The process of calculating sentence weights is optimized by considering sentence features such as sentence position, similarity to the title, keyword coverage rate, and Mongolian conjunctions.Finally, the weight of each sentence is obtained through algorithm iteration.After sorting the sentences, the top two are selected for the summary.Experimental results show that, compared with the TextRank algorithm, the Rouge-1, Rouge-2, and Rouge-L indicators of the proposed method have improved by 0.183, 0.179, and 0.199, respectively.Consequently, the quality of the generated Mongolian summarization is enhanced.