Prototipo de descripción de figuras geométricas para no videntes : realizar un pdf inclusivo, donde conste la descripción de figuras obtenidas de un archivo latex usando un LLM.

Abstract

This study details the development of a web application designed to create inclusive PDFs, aimed at enhancing the accessibility of figures in SVG and TikZ formats for individuals with visual impairments. The initiative addresses the critical need for providing detailed textual descriptions alongside images in scientific, educational, and technical documents, thereby facilitating access to visually represented information. The application employs a Large Language Model (LLM) to analyze and describe complex figures, integrating these descriptions as alternative text tags within the original LaTeX code. This process is automated, ensuring accuracy and consistency in the descriptions provided. Additionally, technologies such as the Web Speech API and gTTS are implemented to offer an auditory experience, allowing users to receive audio feedback while navigating the application. The web application also supports keyboard navigation and offers variable reading speeds to accommodate user preferences. The use of the Flask framework and the Kanban methodology in project development ensured agile and efficient management, facilitating the implementation and testing phases of the system. The results indicate that the application is effective in producing accessible documents, significantly enhancing digital inclusion for individuals with visual impairments. This tool promotes greater equity in access to information, contributing to the broader goal of a more inclusive society. The integration of IA and speech technologies underscores the potential for innovative solutions in addressing accessibility challenges in digital content.