Search papers, labs, and topics across Lattice.
This paper introduces DocMaster, a hierarchical structure-aware system designed to enhance document analysis by preserving the original layouts of complex documents such as academic papers and technical manuals. By constructing hierarchical document trees and a structure-aware semantic index, DocMaster improves the accuracy of document filtering and enables more effective in-depth analysis through natural language queries. The system's performance is validated via an interactive web interface that allows users to seamlessly upload collections, filter documents, and conduct question answering, showcasing significant improvements over traditional flat-text approaches.
Hierarchical structures in documents can boost analysis accuracy by preserving context, leading to more effective filtering and question answering.
Leveraging large language models (LLMs) to analyze complex documents -- such as academic papers, technical manuals, and financial reports -- has emerged as a mainstream and critical task in both research and industry. In practice, users must first filter relevant documents from large collections and then conduct in-depth analysis (e.g. question answering) over the selected subset, yet existing systems flatten documents into plain-text chunks, discarding the rich hierarchical structures (sections, tables, figures, equations) and degrading downstream performance. We present DocMaster, a hierarchical structure-aware document analysis system. DocMaster parses documents into hierarchical document trees preserving original layouts and constructs a structure-aware semantic index that enables accurate document filtering and in-depth analysis. We demonstrate DocMaster through an interactive web interface that enables users to upload document collections, construct tree-based and multi-view semantic indices, filter relevant documents via natural-language conditions, and perform follow-up question answering over the filtered results. The source code, data, and demo are available at https://doc-master.github.io/.
