Slidebot: Automated Presentation Generation Using NLP

Abstract

NLP SlideBot is a fully developed and advanced system designed to automatically generate professional PowerPoint presentations from unstructured text documents. It utilizes modern Natural Language Processing (NLP) methods to transform unstructured text into well-structured, coherent, and visually appealing slides. At its core, NLP SlideBot employs a hybrid summarization model that combines powerful abstractive models like BART[2] and T5[3] with expressive techniques such as TextRank[1] to extract and organize essential information. This process is further supported by BERT-based semantic segmentation[4], which effectively groups related content to form logical slide topics. In addition to text analysis, the system offers several AI-enhanced features, including interactive knowledge graph visualization, automated speaker notes, text-to-speech audio narration, and image generation using Google Gemini[10]. The architecture of NLP SlideBot is built on a current, modular tech stack with a React-based frontend[9] and a FastAPI[8] backend. It supports a range of input formats, including PDF, DOCX, and images, with OCR functionality for text extraction from images. Performance evaluations on a large scale demonstrate that NLP SlideBot is an efficient tool applicable across corporate, academic, and research environments, significantly reducing the time required to create presentations while maintaining high standards of visual design and content accuracy.