Design And Implementation Of An Efficient Text Summarization Method Using Self-Organizing Maps
Text summarization helps with obtaining the significant parts or the core meaning of a piece of text without losing any important message conveyed by the text.In this paper, a method to summarize the text in the English language has been proposed. The proposed method performs extractive summarization using clustering by employing a deep neural network known as the self-organizing map. The text data is divided into sentences, then processed to obtain bare text without any punctuations, and then it is further processed to remove meaningless transformations corresponding to the way the word appears in the sentence, and then they are given scores based on some properties. Then, a transformer-based pre-training network called “MPNet” is used to obtain sentence embeddings for the processed sentences. Afterward, the sentence embeddings are used as input to a self-organizing map of a size determined by the input of the user. Once the self-organizing map outputs the clusters, the sentences are chosen from various clusters based on their corresponding scores. The chosen sentences will then be used to generate the summary.
Keywords - Extractive Text summarization, Unsupervised Text summarization, Self-Organizing Maps, MPNet, Unsupervised Learning, Machine Learning, Deep Learning, Natural Language Processing.