Telling Stories from Computational Notebooks: AI-Assisted Presentation Slides Creation for Presenting Data Science Work
Project Overview
The document explores the use of generative AI in education through the development and evaluation of NB2Slides, an AI-assisted tool that aids data scientists in creating presentation slides from computational notebooks. By leveraging deep learning and example-based prompts, NB2Slides addresses common challenges faced by data scientists, including the difficulties of locating key information, organizing narratives effectively, and customizing content for various audiences. User evaluations reveal that the system significantly enhances efficiency and facilitates content generation, yet highlights the necessity of human intervention to maintain the relevance and quality of the presentations produced. Overall, the findings suggest that while generative AI tools like NB2Slides can streamline educational processes, the role of human oversight remains essential in achieving optimal outcomes.
Key Applications
NB2Slides
Context: Data science projects requiring presentation of findings to stakeholders, including both technical and non-technical audiences.
Implementation: NB2Slides integrates with Jupyter Lab and takes users' notebooks as input to generate draft slides using deep learning algorithms and example-based prompts.
Outcomes: Participants reported improved efficiency in creating slides, reduced complexity in the process, and better organization of content tailored to audience needs.
Challenges: Participants expressed concerns about the accuracy of automated content and the necessity for human refinement to add business context and narrative.
Implementation Barriers
Technical barrier
The AI system may struggle with messy or poorly documented input notebooks, leading to inaccuracies in slide content generation.
Proposed Solutions: Users are encouraged to clean and document their notebooks better before using NB2Slides to enhance output quality.
Trust barrier
Users expressed skepticism about fully automated solutions, feeling that human intervention is necessary for quality and context.
Proposed Solutions: Promoting a human-AI collaboration model where automation aids but does not replace human input.
Project Team
Chengbo Zheng
Researcher
Dakuo Wang
Researcher
April Yi Wang
Researcher
Xiaojuan Ma
Researcher
Contact Information
For information about the paper, please contact the authors.
Authors: Chengbo Zheng, Dakuo Wang, April Yi Wang, Xiaojuan Ma
Source Publication: View Original PaperLink opens in a new window
Project Contact: Dr. Jianhua Yang
LLM Model Version: gpt-4o-mini-2024-07-18
Analysis Provider: Openai