A new Data Visualisation Paradigm By Leveraging An AI Agent
Microsoft Research Is Revolutionising Data Visualisation with Concept Binding
A key element is that as a user, you can describe the analysis you would like to perform, and by leveraging a Language Model, the data analysis is performed.
Most modern visualisation tools require authors to transform data into tidy formats to create desired visualisations.
Data transformation is a barrier in visualisation authoring due to the need for programming experience or separate data processing tools.
Concept binding is a new visualisation paradigm that separates high-level visualisation intents from low-level data transformation steps, utilising an AI Agent.
Data Formulator is an interactive visualisation authoring tool that implements the concept binding paradigm.
In Data Formulator, authors define data concepts for visualisation using natural language or examples and then bind them to visual channels.
Data Formulator’s AI Agent automatically transforms input data to surface defined concepts and generate desired visualisations.
Data Formulator presents results (transformed table and output visualisations) from the AI Agent with feedback to help authors inspect and understand them.
A user study with 10 participants demonstrates that they could learn and use Data Formulator to create visualisations involving challenging data transformations.
Above, the Data Formulator UI. After loading the input data, the authors interact with Data Formulator in four steps:
- In the Concept Shelf, create (e.g., Seattle and Atlanta) or derive (e.g., Difference, Warmer) new data concepts they plan to visualise.
- Encode data concepts to visual channels of a chart using Chart Builder and formulate the chart.
- Inspect the derived data automatically generated by Data Formulator, and
- Examine and save generated visualisations. Throughout the process, Data Formulator provides feedback to help authors understand generated data and visualisations.
Run Data Formulator Locally
First, within your terminal window, from the command line, create a virtual environment called data_formulator.
python3 -m venv data_formulatorThen activate the virtual environment in the following way:
source data_formulator/bin/activateInstall data_formulator …
pip install data_formulator And then run data_formulator on a port which is not in use…
data_formulator --port 5001Data Formulator will then be available on the URL: http://localhost:5001/.
As seen in the video below, the first thing to do, is to select the Language Model you will use for Data Formulator.
With the OpenAI detail added…
Chief Evangelist @ Kore.ai | I’m passionate about exploring the intersection of AI and language. From Language Models, AI Agents to Agentic Applications, Development Frameworks & Data-Centric Productivity Tools, I share insights and ideas on how these technologies are shaping the future.
