Data Story Generation Using Beam Search for Coherent Narrative
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional data narrative systems are limited in generating coherent and comprehensive data stories, often focusing on narrow perspectives and requiring extensive user input and analysis to create visually organized stories, which can be inefficient and fail to uncover anticipated phenomena.
Innovation Solution
A data story generation system utilizing a beam search algorithm to extract important facts from a dataset, generate a sequence of facts, and connect them with coherence facts, providing a global perspective and coherent sequence of visualizations and captions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If conventional data narrative systems are used to generate data stories, then visualizations can be created to reflect trends, but the system requires extensive user input and analysis, reducing efficiency
Solution Approach 1:
The system automatically generates data stories by selecting and organizing visualizations based on dataset characteristics without requiring extensive user input. The automated selection process enables the system to serve itself in creating coherent narratives, improving productivity while reducing operational complexity
Solution Approach 2:
The system changes the parameter of user interaction from extensive manual input to minimal guidance. By automating the visualization selection and story construction processes, the system transforms the operational mode while maintaining output quality, resolving the contradiction between efficiency and ease of operation
2Adaptability or versatility
If conventional data narrative systems generate visualizations, then key metrics can be illustrated, but the system focuses on narrow perspectives and fails to uncover anticipated phenomena
Solution Approach 1:
The system adds a new dimension to data story generation by incorporating diverse visualization types and multiple analytical perspectives simultaneously. This dimensional expansion allows the system to cover broader perspectives and discover unanticipated phenomena that single-perspective approaches would miss
Solution Approach 2:
The system segments the data analysis process into multiple independent visualization selections that can be combined into comprehensive stories. By dividing the narrative into modular components with different perspectives, the system achieves broader coverage while maintaining coherent storytelling
3Reliability
If users compile series of data visualizations manually, then coherent stories can be created, but the process is inefficient and time-consuming
Solution Approach 1:
The system performs preliminary actions by pre-selecting and organizing visualizations based on dataset characteristics before final story assembly. This advance preparation maintains story coherence while significantly reducing the time required for manual compilation and arrangement
Solution Approach 2:
The system introduces an intermediary automated selection process that bridges raw data and final coherent narratives. This intermediary layer maintains story quality by intelligently selecting and ordering visualizations, while eliminating the time-consuming manual compilation process
Data Source
AI summary
Methods, systems, and non-transitory computer readable media are disclosed for intelligently generating comprehensive and relevant visual data stories from tabular data. The disclosed system can create data stories from a given dataset guided by some user selections to set a theme for the story. In particular, the disclosed system evaluates facts of a dataset to identify significant facts. The disclosed system can order significant facts to create a backbone for the data story based on user selections by utilizing a beam search. Furthermore, the disclosed system can expand the backbone of the data story by generating coherence facts. The disclosed system may present the generated story via a graphical user interface.