A Comparative Study of Sentence-Level Techniques for Single-Document Dataset Text Summary Generation

Abstract

Obtaining relevant information from the expanding volume of internet data can pose a challenge for users. To address this issue, Automated Document Summarization has gained significant attention. The key component of a successful document summarizer is a robust document representation. The field of information systems research focuses on automatic text summarization, which aims to generate concise and informative summaries without requiring specialized knowledge. Extractive text summarization, one approach in this field, involves ranking sentences based on various factors. Various researchers have proposed several feature-based scoring algorithms to support this technique. Text summarization plays a vital role in automatically condensing text documents, facilitating the discovery of pertinent details from the Internet or vast libraries. In this paper evaluated various sentence features method for extractive text summarization for single document text summarization. The important sentences are selected using text feature extraction methods such as TF-IDF, Thematic Feature, and Word Frequency. These sentences are then sorted based on their scores, and final document summaries are generated. The generated summaries are compared with golden summaries from a dataset, and the evaluation is performed using ROUGE scores and BLEU. Additionally, the paper provides an overview of methodologies and a comparative table of approaches used by researchers for evaluation and advancement in the field of text summarization.