Word Spectral Visualization Base on Fruchterman-Reingold Optimized ForceAtlas2

Abstract

This article presents a visualization technique employing the BERT model similarity lexicon and delves into the method of picking and enhancing layout algorithms when generating similar term lexicon exhibitions. In a bid to achieve superior visualization outcomes, two distinct layout algorithms, namely ForceAtlas2 and Fruchterman-Reingold, are utilized to leverage their individual merits and augment visualization quality. The similar term lexicon's visual depiction is accomplished using the ForceAtlas2 and Fruchterman-Reingold layout algorithms, with the amalgamation of ForceAtlas2's initial layout and Fruchterman-Reingold's fine-tuning resulting in a presentation map of the similar term lexicon with a superior visualization impact. This tool equips us with an efficient method to profoundly interpret the semantic relationships of textual data. Through this method, we can perceive the connections and associations among words more explicitly, aiding in unveiling the concealed information and meanings in the text data. Further research can further refine the visualization and broaden the application domain to cater to the escalating demands. This method is anticipated to play a vital role in text-connected tasks, offering a potent tool for extracting knowledge and information from textual data