🤯 Teaser
Ever wanted to get a complete analysis of a situation so you can take data-driven decisions for any domain and achieve data-analysis...within a few seconds ?
🫵 Well...I have good news for you, this short post is all about that.
🍿 For the inpatients
❔ Inception
Everything started with the following tweet:
... talking about a
"Systems that support users in the automatic creation of visualizations, [...] understand the semantics of data, enumerate relevant visualization goals and generate visualization specifications."
I was just wondering how far and how good this could be so I wanted to give it a try on:
- A very concrete real life use case,
- A public Notebook on a on a public Kaggle Dataset (🥷 Neo4J Ninjas duckdb dataset 🦆)
📜 The analysis workflow
The analysis workflow is pretty straightforward and can be splitted into two main phasis: preparation then get insights:
- 🗂️ Pick a dataset on which your business depends
- 🧍 Design a persona (ie. to modelize the business case)
- 🎯 Ask for a number of goals
- 🎯 Get goals ideas (and detailed explanations) from AI
- 📊 Get dataviz from AI
-
🧠 Analyze : interact with
LIDA
in a continuous improvement loop (aka. use your own brain to tune persona, dataset,... 😏)
🦥 Enjoy goals... as data
It's as simple as setting a number of goals and the target persona:
goals = lida.goals(summary,
n=5,# Let's get 5 goals
persona="Neo4J community manager who
needs to manage its worldwide community to get the best
community engagement.")
Then within less than a second you get the goals (eventually as a pandas
dataframe so you can also share goals as a spreadsheet):
🥳 Finally you just have to ask for some dataviz for any goal:
📑 Resources
- 🤖😘🥷 Neo4J community manager automation by
LIDA
- https://aclanthology.org/2023.acl-demo.11/
- https://github.com/microsoft/lida
- https://microsoft.github.io/lida/
- LIDA | Automatically Generate Visualization with LLMs | The Future of Data Visualization
- An Overview of LIDA: Generate Visualizations and Infographics of Tabular Data using LLMs!