FactFlow: Automatic Fact Sheet Generation and Customization from Tabular Dataset via AI Chain Design & Implementation

TL;DR

This paper tackles the challenge of generating informative fact sheets from tabular data for non-experts by introducing FactFlow, an AI-chain system that assembles five specialized AI workers to produce interconnected facts, visuals, and narrative structure. It defines a taxonomy for configuring AI workers, develops a dataset representation and anonymization pipeline, and provides a user-friendly NL-based customization interface. The approach is validated on real-world datasets and through an 18-participant user study showing superior content quality, visual appeal, and customization usability compared with two baselines. The work demonstrates a practical, scalable framework for automated data storytelling that can be deployed as a web platform to support diverse audiences.

Abstract

With the proliferation of data across various domains, there is a critical demand for tools that enable non-experts to derive meaningful insights without deep data analysis skills. To address this need, existing automatic fact sheet generation tools offer heuristic-based solutions to extract facts and generate stories. However, they inadequately grasp the semantics of data and struggle to generate narratives that fully capture the semantics of the dataset or align the fact sheet with specific user needs. Addressing these shortcomings, this paper introduces \tool, a novel tool designed for the automatic generation and customisation of fact sheets. \tool applies the concept of collaborative AI workers to transform raw tabular dataset into comprehensive, visually compelling fact sheets. We define effective taxonomy to profile AI worker for specialised tasks. Furthermore, \tool empowers users to refine these fact sheets through intuitive natural language commands, ensuring the final outputs align closely with individual preferences and requirements. Our user evaluation with 18 participants confirms that \tool not only surpasses state-of-the-art baselines in automated fact sheet production but also provides a positive user experience during customization tasks.

Figures (5)

• Figure 1: Workflow of FactFlow for generating a fact sheet using a global population dataset with columns for Country, Population, and Continent. FactFlow first prepares the dataset representation, which is then processed by the AI Chain. Each worker in the chain is profiled for specific tasks to generate the fact sheet.
• Figure 2: Our FactFlow interface: a) User can upload dataset and provide any additional natural language request to generate the fact sheet b) Based on the initial fact sheet, users can manually add, remove and edit the content, as well as generating a new fact using natural language request.
• Figure 3: Two fact sheets that automatically generated by FactFlow based on real-world datasets.
• Figure 4: Radar charts showing the quality ratings of fact sheets produced by FactFlow compared to the Calliope and ChatGPT Plus baselines, evaluated by three participant groups: N1-N6 (novices), J1-J6 (junior-level data workers), S1-S6 (senior data workers), and ALL (all participants).
• Figure 5: A) Bar chart showing ratings from different user groups on Accuracy, Alignment, Aestheticness for the customized fact sheet. B) Bar chart displaying average SUS feedback from participants on the use of the FactFlow platform.