SUMMPILOT: Bridging Efficiency and Customization for Interactive Summarization System

TL;DR

The paper presents SummPilot, an interactive, LLM-based summarization system that bridges efficiency and customization by combining automatic summarization with interactive components (semantic graphs, entity clustering) and an explainable evaluation loop. It enables multi-document summarization through a graph-based representation and supports user-driven control via Basic and Advanced modes. Two user studies demonstrate high usability and usefulness, with Advanced Mode improving understanding and control, suggesting practical value for professionals and students. Future work targets multilingual and multimodal LLM integration to broaden applicability and improve human–system interaction.

Abstract

This paper incorporates the efficiency of automatic summarization and addresses the challenge of generating personalized summaries tailored to individual users' interests and requirements. To tackle this challenge, we introduce SummPilot, an interaction-based customizable summarization system. SummPilot leverages a large language model to facilitate both automatic and interactive summarization. Users can engage with the system to understand document content and personalize summaries through interactive components such as semantic graphs, entity clustering, and explainable evaluation. Our demo and user studies demonstrate SummPilot's adaptability and usefulness for customizable summarization.

Figures (2)

• Figure 1: Interface design and components for SummPilot (Advanced Mode): Users can explore documents through semantic graphs based on relational triples [A], and manage content via checkboxes for specific triples [B]. Entity clustering groups related expressions [C]. Open-form commands allow custom inputs [D] and the "Evaluate" button provides feedback on compression, coverage, and consistency, helping to refine user control and model performance [E].
• Figure 2: Backend pipeline of SummPilot.