Browsing by Author "PACHECO, PEDRO RAFAEL FERREIRA"
Now showing 1 - 1 of 1
Results Per Page
Sort Options
- Generating presentation documents from procedural resourcesPublication . PACHECO, PEDRO RAFAEL FERREIRA; Azevedo, Isabel de Fátima SilvaThe creation of presentation slides from procedural documents is a repetitive and time consuming task that typically requires substantial manual e ort. This dissertation presents the design and development of a system that automates this process by converting procedural documents into structured and visually coherent presentation slides. The proposed solution integrates Natural Language Processing (NLP) techniques, text and image extraction and the use of local Large Language Model (LLM)s to perform summarization and content structuring. The system follows a modular pipeline composed of three main stages, extraction of textual and visual elements from Portable Document Format (PDF) documents, processing and structuring of the extracted content using LLMs based summarization and organization strategies and automatic generation of presentation slides in PowerPoint Open XML Presentation (PPTX) format, enriched with customizable themes. The implementation leverages open-source tools such as PyPDF2, pdfminer.six and python-pptx, while ensuring exible integration with di erent LLMs through the Ollama framework. Evaluation was carried out through unit, integration, functional and non-functional testing. The results con rmed that the system ful lls its functional requirements, reliably importing documents, generating coherent slides, supporting customization and export features. Nonfunctional testing highlighted strengths in usability and maintainability, while also exposing limitations in performance when using larger models, reliability under heavy load and the need for improved security hardening. This work proves the concept of value and feasibility in semi-automatic transformation of procedural documents into presentations. The proposed system reduces the time and e ort required for preparing presentations, o ering an accessible, modular and extensible solution that can be further optimized and scaled in future research. The source code is publicly available as open source for testing, improvement and continued development (Pacheco, 2025).