Digilog: Enhancing Website Embedding on Local Governments - A Comparative Analysis

The ability to understand and process websites, known as website embedding, is crucial across various domains. It lays the foundation for machine understanding of websites. Specifically, website embedding proves invaluable when monitoring local government websites within the context of digital transformation. In this paper, we present a comparison of different state-of-the-art website embedding methods and their capability of creating a reasonable website embedding for our specific task based on different clustering scores. The models consist of visual, mixed, and textual-based embedding methods. We compare the models with a base line model which embeds the header section of a website. We measure their performance in an off-the-shelf evaluation as well as after transfer learning. Additionally, We evaluate the models’ capability of distinguishing municipality websites from other websites such as tourist websites. We found that when taking an off-the-shelf model, Homepage2Vec, a combination of visual and textual embedding, performs best. When applying transferred learning, MarkupLM, a markup language-based model, outperforms the others in both cluster scoring as well as precision and F1-score in the classification task. All mixed or markup language-based models achieve an F1-score and a precision over 97%. However, time is an important factor when it comes to calculations on large data quantities. Thus, when additionally considering the time needed, our base line model performs best.