Novel Similarity Analysis and Comparison System

Abstract

This research presents the application of Similarity Analysis techniques for analyzing and comparing novels in both Thai and English. The system supports importing novels as single or multiple files in TXT, DOCX, and PDF formats, as well as direct copy-and-paste text input. The system's workflow begins with preprocessing and preparing the novel data, including text cleaning, formatting, and segmentation into analysis units. The system can perform similarity analysis at multiple levels, ranging from words to sentences, paragraphs, chapters, and entire novels. The results are presented in the form of similarity scores, and various visualizations, including heatmaps, network graphs, and timelines. Performance evaluation conducted on an English novel dataset comprising 50 titles (150 chapters) across 5 genres demonstrated that at an optimal similarity threshold of 0.60, the system achieved a precision of 84.8%, a recall of 82.0%, and an F1-score of 83.4%, indicating high effectiveness and reliability in text similarity detection and classification.