Effectively Delivering Author’s Point to Reader: Pointer-Generator Network Approach

Abstract

We propose a communication-based text summarization approach utilizing a pointer-generator network. The proposed summarization algorithm combines elements of abstractive summarization with keyword extraction. These keywords are associated with threats, causality, evidence, and solutions, all aimed at influencing readers' behavior. This paper is grounded in the theoretical frameworks of Language-Action Perspective (LAP) and Speech Act Theory (SAT), which are applied within the context of social and behavior change communication (SBCC). Consequently, our ultimate model considers the target readers to enhance the effectiveness of communication for promoting social and behavioral change. The SBCC-based summary incorporates more persuasive and emotive language while prioritizing SBCC-related content from the original article. This study demonstrates the potential effectiveness of a summary that encapsulates the author's intentions in shaping readers' thoughts and behaviors.