The Day Mom Learned About Vector Embeddings
Last Sunday afternoon, Mom was rolling out chapatis when she noticed I was staring at my laptop.
Mom: "What are you working on this time? You look like you’re solving a puzzle."
Me: "Kind of! I’m working with something called vector embeddings."
Her eyebrows raised.
Mom: "Sounds like some fancy sewing technique."
And that’s when I knew it was time for a kitchen-table explanation.
A Story About a Magical Spice Rack
I told Mom to imagine we owned a magical spice rack.
In this rack, every recipe — biryani, dal, dosa isn’t just kept in words.
Instead, each one is turned into a secret code a set of numbers that tell where it belongs.
Why not just use recipe names?
I asked her:
“Mom, if you keep dal tadka and chana masala next to each other on the rack, why would you do that?”
She said:
“Because they’re both dals, both spicy, and they use similar masalas.”
Exactly. Vector embeddings do the same thing — they put similar things close together in a special “number world.”
How it works (Biryani Edition 🍚)
Imagine every recipe is turned into a dot on a giant invisible map:
Veg Biryani: near Pulao and Jeera Rice
Dal Tadka: near Chana Masala
Gulab Jamun: far away in the “dessert” section
The numbers that describe the dot’s position are the vector embedding.
Kitchen shelf
Meals Items
🟡 Veg Biryani
🟢 Pulao
🟢 Jeera Rice
Desserts
🔴 Gulab Jamun
The closer two dots are, the more related they are.
Why it’s useful
When a computer has these maps of meaning, it can:
Find similar recipes even if you use different words (“pulao” vs “pilaf”).
Group together songs, news articles, or images that share a theme.
Help search engines understand your question, not just match exact words.
Mom’s reaction
After I explained, Mom smiled:
“So… a vector embedding is just the computer’s way of arranging things so it can find them later?”
Exactly. It’s like a magical spice rack, but instead of turmeric and cumin, it’s made of numbers and meaning.
Conclusion
Vector embeddings aren’t about stitching clothes.
They’re about placing information in a smart, invisible map so computers can understand relationships the same way we just know dal tadka is closer to chana masala than to gulab jamun.
