Use ChatGPT to create structured data from unstructured data

Structured data helps you analyze it, but making it structured can be a pain

Let’s say you’re a Alice the Accountant at LargeTradCorp and Joe the Sales Guy emails you a quarterly report with expected sales numbers from clients he’s close to closing. His email reads:

Hey Alice, how are you? Sorry to hear about your kid’s strep throat—hope she feels better soon—I remember when my son had it last year, he felt sick for what seemed like months! Anyway, so here’s the numbers we talked about last week: Bob’s Oil Services $60,000, Alice’s Funeral Services $85,000 (I think), Jim’s Flower Arrangements $15,000, Dave’s Sculptural Arrangements $75,000. Also, by the way, I’m going to be in the office next week, so I’ll pick up my expense reimbursement then. Thanks, Joe

Great, you think to yourself, now I have to type everything into Excel so that I can actually do something with it. Joe just creates more work for me.

But, it turns out that ChatGPT can take the unstructured data in Joe’s email, and turn it into a beautiful table that you can copy and paste into Excel! Here’s how it works:

Prompt: put this into a markdown table: Bob's Oil services, $60,000; Alice's Funeral services, $85,000; Jim's Flower Arrangements, $15,000; Dave's Sculptural Arrangements, $75,000

Here’s ChatGPT’s output:

You can even add some lightweight analysis to this: Add a sum and an average at the bottom of the table. Here’s the output:

And that’s it! Any time you have unstructured data, such as from Joe’s email, you may be able to get ChatGPT to wrangle it for you into a well-formatted, structured form.

Here’s a Twitter thread full of examples:

James Blackwell @jwblackwell

GPT 3 / ChatGPT is great at turning unstructured data into structured data. Probably the most underrated use case. E.g. Entity extraction:

If you want to learn more about unstructured data, there are a ton of links out there for you. Here’s one. Or we could ask ChatGPT for a definition:

Unstructured data refers to data that does not have a pre-defined data model or does not fit neatly into a database structure. It is information that does not conform to a particular format or structure and cannot be easily organized into a table or database schema. Examples of unstructured data include text documents, email messages, audio and video files, social media posts, and images. Unlike structured data, which can be processed and analyzed using automated methods, unstructured data often requires manual processing to extract meaningful information.

Now that you know what unstructured data is, and that ChatGPT can help you wrangle it into a more usable format, you should be able to identify lots of occasions in your daily work where having structured data would be more useful.