· Balisage Series on Markup Technologies· 0 citations
TL;DR
A “separation of concerns” (between the generative model and the structured data it consults) is proposed as a principled, declarative approach to AI-assisted XML work.
Abstract
This paper reflects on experiments to integrate Small Language Models (SLMs) with XML technologies in academic Digital Humanities work, and discusses the “DigitAI” project. The project is motivated by two goals for making AI systems: 1) more educationally transparent and 2) more economically and environmentally sustainable. We first attempted to make a small local Qwen model serve as an assistant for questions about applying the TEI P5 Guidelines. We provided Qwen with a Retrieval Augmented Generation (RAG) system built from the TEI P5 Guidelines, using XSLT to transform the guidelines into a neo4j knowledge graph. This approach proved both bloated and disappointing: the model's conversion of structured XML data into word embeddings eroded the precision of the original markup, and the outputs of prompts reflected “hallucination” and inaccuracy.
Learning from this failure, we came to realize that XML should be kept as XML, held apart from the language model's internal machinery, and made accessible instead through a Model Context Protocol (MCP) server that allows the SLM to query the TEI document tree directly using XPath and related XML stack technologies. We have been experimenting with MCP scripting and evaluating how well SLMs apply XPath, XSLT, and XQuery parsers, giving them agency to evaluate and contribute to XML and TEI XML projects. Much of our learning process has been in preparing a containerized system with Docker for our project to run safely in controlled conditions on personal computers, and to share with others across platforms. Reflecting on our experience, we propose a “separation of concerns” (between the generative model and the structured data it consults) as a principled, declarative approach to AI-assisted XML work.
Querying LLMs as digital libraries is feasible, but its effectiveness depends on model strength, deployment conditions, dataset structure, and execution strategy, and Galois remains valuable when relational discipline and controlled query execution are required.
Thirty years of working with markup languages, semantic web standards, and knowledge representation systems reveals a single recurring problem wearing many different masks: the problem of bounded, scoped, composable meaning. This paper traces a personal and technical journey from the HTML Document Object Model of 1996 through XML, XSLT, XSD, XQuery, RDF, SHACL, and the neural systems of the present day, arriving at the holonic graph as the resolution of a tension that has persisted, largely unacknowledged, throughout the life of the markup community. The holon — Arthur Koestler's term for a unit that is simultaneously a whole and a part of a larger whole — turns out to have been implicit in every major design decision the XML and semantic web stacks ever made. RDF 1.2, SHACL 1.2, and named graphs now provide the formal apparatus to make it explicit. The implications extend from knowledge governance to the grounding of large language models.
K. Cagle· Balisage Series on Markup Te...· 0 citations
Multilingual documents encapsulate rich regional cultures, scientific discoveries, and historical records. Parsing this content into structured, machine-readable formats is critical for unlocking global knowledge. However, existing benchmarks predominantly focus on high-resource languages like English and Chinese, creating an evaluation blind spot concerning model performance on other languages. While recent Vision-Language Models (VLMs) claim support for hundreds of languages, the lack of ground truth makes it impossible to empirically verify these capabilities. To bridge this gap, we introduce MORE, a large-scale benchmark designed for multilingual document parsing evaluation. MORE distinguishes itself through three key dimensions: (1) Unprecedented Scale: It covers 149 languages, making it the most linguistically diverse benchmark to date; (2) Structural Complexity: Unlike previous works, it extends evaluation beyond plain text to include structural elements such as code blocks, tables, and catalogs; and (3) Data Authenticity: All samples are curated from real-world documents via a model-assisted, human-refined annotation pipeline. We evaluate state-of-the-art models using MORE, establishing new performance baselines for long-tail languages and validating the benchmark's effectiveness in diagnosing model capabilities in realistic, diverse scenarios. The MORE dataset will be available at https://github.com/zimoqingfeng/MORE.
Long Xu, Binghong Wu, Ting-Ting Yu et al.· 3 citations
Documents remain one of the primary carriers of knowledge in scientific, enterprise, and governmental settings, yet their complex visual structure—comprising layouts, tables, figures, and multi-column text—poses significant challenges for modern AI systems. Large language models do not natively understand such document structures, while traditional PDF parsers often lose semantic and structural information, leading to noisy or incomplete representations. In this talk, we present Docling, an open-source document processing framework designed to transform heterogeneous documents into high-quality, structured representations suitable for AI applications. Docling combines layout analysis, OCR, table structure recognition, and document assembly into a unified pipeline that produces rich, LLM-friendly outputs such as structured JSON and Markdown while preserving visual grounding and reading order. The framework supports multiple input formats and integrates with modern AI ecosystems, enabling applications including retrieval-augmented generation, schema-based information extraction, and agent-driven workflows. We discuss the architecture of Docling, its emerging ecosystem of models and tools, and its role in enabling scalable, privacy-preserving document intelligence through local execution and open standards. By bridging the gap between visually complex documents and machine-readable knowledge, Docling provides a foundation for reliable document understanding in next-generation AI systems.
P. Staar· Proceedings of the 2026 ACM...· 0 citations
Academic papers are a primary carrier of scientific knowledge, yet most of this knowledge remains locked in PDFs that are optimized for human reading rather than machine use. For Multimodal Large Language Models (MLLMs), the core challenge is not only perception, but representation: scientific pages interleave text with Structured Academic Elements (SAEs) such as tables, formulas, charts, and pseudocode, whose structure, data, and logic are poorly preserved by common surrogates like Markdown. We therefore propose Compilable Academic Document Parsing (CADP), a paradigm that reconstructs a full page as contextual \LaTeX{} plus executable Python, so that structure-preserving elements and executable chart representations can be reconstructed, recompiled, and directly verified against the source page. To support this setting, we introduce CADP-Bench, an expert-verified benchmark of full academic pages containing tightly coupled text and multiple SAE types, evaluated through a re-injection compilation protocol. We further study current capabilities using SOTA MLLMs and an exploratory multi-agent baseline that incorporates common agentic techniques. Results show that even frontier models still struggle to produce high-fidelity executable reconstructions, highlighting substantial room for improvement in structure-aware scientific document parsing. CADP-Bench is released for future research.
This paper presents the design and implementation of intelligent chatbots using Retrieval-Augmented Generation architecture to derive accurate, context-based responses from domain-specific knowledge sources such as PDFs and CSV files. The proposed system integrates information retrieval mechanisms with large language models to overcome crucial limitations of traditional chatbots: hallucination, knowledge cut-off, and inability to access private organizational data. The proposed architecture uses a six-stage pipeline consisting of document ingestion, text splitting, and embedding generation with sentence transformers, semantic retrieval, and response generation via Large Language Models (LLMs). By preprocessing the PDF and CSV documents using a RecursiveCharacterTextSplitter and converting them into 384-dimensional semantic embeddings, the system works via a similarity-based retrieval of document chunks. For implementation, the LlamaIndex framework was used for data indexing and query orchestration, and HuggingFace Transformers for language modeling. Experimental validation conducted on the smart campus chatbot system demonstrated significant improvements in the accuracy, relevance and transparency of responses across education, student services, and administrative support domains compared with a standalone language model. Because of its modular design, it supports a wide range of domain-specific topics without the combinatorial costs of re-training. This is another important application for systems that require regular updates and accurate information. Overall, the proposed framework provides a simple and practical approach for easily implementing domain-aware chatbots.
Varsha Spoorthi Shree Narravula, V. K, Yashaswini B. Suresh et al.· 2026 7th International Confe...· 0 citations