Sistema multiagente para lectura y comprensión de documentos con inteligencia artificial
Loading...
Date
Authors
Journal Title
Journal ISSN
Volume Title
Publisher
Universidad Nacional Arturo Jauretche. Instituto de Ingeniería y Agronomía
Abstract
La acelerada producción de artículos científicos plantea un desafío significativo para la comunidad, limitando la capacidad de asimilación de nueva información. Por este motivo, se desarrolla un Sistema Neuro-Analítico Multiagente con el propósito de complementar y acelerar el análisis automatizado de documentos científicos.
El enfoque metodológico principal se basa en una arquitectura distribuida de microservicios que implementa la Recuperación Aumentada (RAG) y utiliza Modelos de Lenguaje Grande (LLMs). El sistema se organiza en tres planos funcionales: el Plano de Orquestación Conversacional, el Plano de Procesamiento Documental y el Plano de Persistencia y Estado.
Los resultados clave incluyen un sistema capaz de realizar la carga y procesamiento completo de documentos, indexación semántica, recuperación vectorial (RAG), generación de resúmenes contextuales e interacción conversacional fluida. Además, se incorpora la funcionalidad de Reconocimiento Óptico de Caracteres (OCR) tradicional y multimodal.
Como conclusión, se ha logrado desarrollar un sistema robusto que facilita la organización, generación de relaciones semánticas y síntesis de literatura científica mediante una arquitectura modular y escalable. Esto permite la consulta contextualizada y el análisis inteligente de papers, proveyendo una herramienta eficiente para la asimilación de conocimiento.
The rapid production of scientific articles poses a significant challenge to the community, limiting the capacity to assimilate new information. For this reason, a Multi-Agent NeuroAnalytical System is being developed to complement and accelerate the automated analysis of scientific documents. The main methodological approach is based on a distributed microservices architecture that implements Augmented Retrieval (RAG) and utilizes Large Language Models (LLMs). The system is organized into three functional planes: the Conversational Orchestration Plane, the Document Processing Plane, and the Persistence and State Plane. Key results include a system capable of performing complete document loading and processing, semantic indexing, vector retrieval (RAG), contextual summary generation, and fluid conversational interaction. Furthermore, traditional and multimodal Optical Character Recognition (OCR) functionality is incorporated. In conclusion, a robust system has been developed that facilitates the organization, generation of semantic relationships, and synthesis of scientific literature through a modular and scalable architecture. This enables contextualized search and intelligent analysis of papers, providing an efficient tool for knowledge assimilation.
The rapid production of scientific articles poses a significant challenge to the community, limiting the capacity to assimilate new information. For this reason, a Multi-Agent NeuroAnalytical System is being developed to complement and accelerate the automated analysis of scientific documents. The main methodological approach is based on a distributed microservices architecture that implements Augmented Retrieval (RAG) and utilizes Large Language Models (LLMs). The system is organized into three functional planes: the Conversational Orchestration Plane, the Document Processing Plane, and the Persistence and State Plane. Key results include a system capable of performing complete document loading and processing, semantic indexing, vector retrieval (RAG), contextual summary generation, and fluid conversational interaction. Furthermore, traditional and multimodal Optical Character Recognition (OCR) functionality is incorporated. In conclusion, a robust system has been developed that facilitates the organization, generation of semantic relationships, and synthesis of scientific literature through a modular and scalable architecture. This enables contextualized search and intelligent analysis of papers, providing an efficient tool for knowledge assimilation.
Description
Citation
Serapio, A. N. (2025). Sistema multiagente para lectura y comprensión de documentos con inteligencia artificial [Práctica Profesional Supervisada, Universidad Nacional Arturo Jauretche]. https://rid.unaj.edu.ar/handle/123456789/3617