KNOWLEDGE GRAPH-BASED MODELING OF WEB APPLICATIONS

Abstract

Traditional web crawling and parsing techniques, which rely on extracting and recursively following hyperlinks to map a web application, are increasingly inadequate for modern, dynamic web applications. This conventional method essentially creates a graph where nodes are merely chunks of unstructured data (the web page content) and edges represent simple page transitions. It fundamentally fails to capture the rich interactive behaviors and state changes that define contemporary applications. In response, a novel approach is proposed that models the web application not as a collection of links, but as a system of structured states. Under this paradigm, each node represents a precise, organized description of the application's current condition, while the edges strictly denote user-initiated actions or purposeful transitions. This fundamental shift from unstructured content to structured representation yields a far more complete and functional understanding of the application's true behavior. This enhanced model promises significant utility for advanced downstream tasks, particularly in automated testing and sophisticated behavior analysis.