Data Formulator
lastmod 2024-11-11created 2024-11-111 edits
🪄 Create rich visualizations with AI
Features
- A Microsoft Research application that uses LLMs to transform data and speed up data visualization work
- Helps analysts transform data with AI and create rich visualizations iteratively
- Unlike most chat-based AI tools where users must describe everything in natural language, Data Formulator combines user interface interactions (UI) and natural language (NL) input for easier interaction
- Can be installed with Python PIP or used in GitHub Codespaces
- Basic usage
- Provide your own OpenAI key, choose a model (GPT-4o recommended), then select a dataset
- Choose a chart type, then drag and drop data fields onto chart properties (x, y, color, ...) to specify the visual encoding
- You can type field names that don't exist in the current data into the Encoding shelf
- This tells Data Formulator to create a visualization that requires computing or transforming the existing data
- Optionally, you can provide a natural language prompt to explain and clarify your intent (not needed if the field names are self-explanatory)
- Click the Formulate button
- Data Formulator transforms the data and instantiates the visualization based on the encoding and prompt
- Inspect the data, chart, and code
- To create a new chart based on an existing one, follow up in natural language
- Provide a follow-up prompt (e.g., "show only the top 5!")
- You can also update the visual encoding for the new chart