HTML 클러스터링과 심층 언어 모델을 활용한 전자책 템플릿 추천

초록

This study proposes an HTML-based e-book template recommendation system to address the challenges of selecting appropriate templates in e-book production. Traditional e-book templates determine layouts based on the type and format of the content; however, editors with limited e-book production experience often struggle to choose effective templates that align with the characteristics of content. In this study, we developed a system that integrates similarity-based clustering techniques with a deep language model to recommend e-book templates tailored to manuscript features. First, templates were clustered based on structural, stylistic, and manuscript-length similarities in HTML pages. Subsequently, a deep language model was employed to recommend templates best suited to the content characteristics of the manuscript. Experiments using datasets from the e-book publishing industry demonstrated the effectiveness of the proposed recommendation system in addressing real-world challenges in e-book production.

키워드