New Conversational AI Tool Lets You “Chat” With Your Data

As an ML engineer, one area where I spend a lot of time is data engineering. Can we use conversational AI technologies like ChatGPT to help boost productivity here?

Turns out Akkio recently shipped an amazing tool called “Chat Data Prep” that lets you “chat” with your data. That’s right! You can transform your data by simply using instructions. 

Data Transformation 

When preparing data to train an ML model, you need to process, transform and clean it. What if I told you that you don’t need SQL or Python to do this?

Akkio ’s new exciting feature “Chat Data Prep” uses the power of conversational AI to interact and transform your data just using text instructions. Let’s check out a short example.

Let’s say we want to train a sentiment classifier on a restaurant reviews dataset. Here are two transformations you can perform on the text data:

• Removing punctuations
• Lowercase text

Chat Data Prep is brilliant for this! You can transform the data with natural language instead of using code.

Here is a view of how the loaded data looks on the interface, including the columns and rows of reviews and sentiment labels:

Data is never perfect, especially text. You should expect to do a lot of cleaning. I like the idea of just using instructions to get this job done. That’s where Chat Data Prep comes in.

As a quick example, here is how I instructed Chat Data Prep to remove punctuations:

I am also interested in converting all the text into lowercase, which is common when processing NLP datasets. Here is how to easily do it with one text command via Chat Data Prep:

This a great feature because you can just continue chaining all your data transformations.

And you can do a lot more! Here is a demo of two more popular and useful data transformations you can do with Chat Data Prep:

• 3 sigma outlier removal
• Date reformatting

Train A Sentiment Classification Model

One really cool feature I particularly like about Akkio is the ability to quickly train a model to test the impact of the data transformations. Again, no code is needed!

I can see that the transformations I applied boosted the performance of the model. And I can easily go back and continue adding more data transformations to see if I get better performance without leaving the interface. ML is iterative so this is really useful!

Akkio also has useful integrations like Snowflake, BigQuery, Google Sheets, and others. These are useful as you can easily access your data stores and start finding value in your data.

You can even deploy a web app or production-ready API. This really accelerates experimentation and deployment.

Last Words

I get super excited to see how AI-powered tools like Chat Data Prep can boost productivity.

No-code tools are redefining what it is to build and deploy ML-powered apps. Excited about this partnership with Akkio and what’s ahead! Check them out here: Akkio