In many professions, educated professionals have always needed to stay up-to-date with advances in their field. With the increasing number of published research works, finding relevant content has become more challenging. Traditional search methods often return thousands of papers related to a search query, necessitating significant time for full-text analysis. Conversely, existing AI-based systems also fail to provide accurate responses. To address these challenges, this position paper proposes a system architecture for scientific understanding, based on user preferences identified in a preliminary survey, that leverages a scientific knowledge graph extending beyond abstract analysis to capture the broader context of scholarly works. The framework aims to deliver context-sensitive responses to user queries and recommend relevant academic studies, thereby improving comprehension. Its development entails tackling various considerations, which will also be briefly discussed.