Automatic Visualizations with Microsoft LIDA
Microsoft LIDA automatically generates visualizations and infographics from data without needing you to do it yourself. It uses Large Language Models (LLM) and Image Generation Models (IGM) for this. You can use LIDA with any programming language or visualization rules.
LIDA can do things like making a summary of data, figuring out what visualizations you want, creating visualizations, making infographics, explaining visualizations, and suggesting other visualizations.
Here are some specific capabilities of LIDA:
- Data Summarization: LIDA efficiently provides a comprehensive overview of your data, even when it is intricate or disorganized.
- Visualization Exploration: LIDA assists in determining the optimal visualization methods for your data, aligning with your specific objectives and requirements.
- Automated Visualization Generation: LIDA can autonomously generate a diverse range of visualizations, including charts, graphs, and maps.
- Infographic Creation: LIDA is capable of producing infographics, combining visuals and text to communicate intricate information clearly and concisely.
- Interpretation of Visualizations: LIDA aids in comprehending the significance of your visualizations, empowering you to make informed decisions based on your data.
- Visualization Recommendations: LIDA suggests additional visualizations that may prove beneficial, taking into account your data and objectives.
Architecture of LIDA:
- SUMMARIZER takes your dataset and turns it into a short, easy-to-understand summary in plain language. It also adds extra details to help you make sense of the data.
- GOAL EXPLORER uses the summary to create goals for exploring your data. These goals include questions to ask, visualizations to create, and reasons for doing so.
- VISGENERATOR is where the visualizations are actually made. It uses code templates, large language models, and your goals to create visualizations and code for them.
- INFOGRAPHER adds style to your visualizations to make them look good and be easy to understand. It uses text descriptions to create different visual styles that fit your needs.
About Dataset:
Data is about location reviews and ratings of Walmart stores in the USA region.
The data has been collected from ScrapeHero, one of the leading web-scraping services in the world. Click here for the Data Source that we used for analysis!
Columns:
ID, Name, Address, Street, Zip_Code, State, City, Author, Review, Rating
Getting Started with LIDA:
The easiest way to explore the library is to fire up the UI or Web App:
First, export the OpenAI key in the terminal. Since we will be using OpenAI for the blog.
You can find other supported models here.
export OPENAI_API_KEY=”YOUR_API_KEY”
To access UI:
lida ui — port=8080 — docs
Go to localhost:8080 in your browser to access the User Interface(UI) .
Click on demo to interact and visualize your data.
Drag and drop your own data to see the magic of LIDA. By default seaborn will be the visualization library used. You can change it via the dropdown menu.
Then there is Generation Settings which we can use to change the Model Provider, Model, Model settings.
Once we upload our data, the magic of LIDA happens.
First it generates the summary of the data. It provides the summary of the data, datatype, unique values, description of the features etc.
Next step, It creates goals for exploring your data.
Then finally it creates the visualizations based on the goals described or you can also pass custom queries to see visualizations.
Now let’s see some of the example queries and visualizations:
Query: What is the overall rating distribution of Walmart stores?
Query: What is the overall sentiment of the Florida region?
Query: What are the common words used in the customer reviews?
It will also provide you recommendations based on the dataset.
Conclusion:
LIDA marks a notable advancement in both the accessibility and capability of data visualization.
By streamlining the entire journey from interpreting data to refining visuals, it enables individuals with varying technical expertise to reveal the narratives concealed within their data.
Whether you’re an experienced data scientist or an inquisitive beginner, LIDA transforms into your dedicated data visualization assistant, leading you through each stage and providing an extensive array of tools.
Hope you learned something new today, Happy Learning!