Python Slot Filling and Extraction — ELI5

Imagine you call a pizza place and say “I’d like a large pepperoni pizza delivered to 42 Oak Street.” The person on the phone grabs three pieces of information: size (large), topping (pepperoni), and address (42 Oak Street). They fill in an order form, one blank at a time.

Slot filling is how a chatbot does the same thing. When you send a message, the bot looks for specific details — the “slots” — and drops them into the right blanks on its invisible form.

The slots depend on what the bot does. A travel bot needs destination, date, and number of travelers. A banking bot needs account type and amount. A weather bot just needs a city name. Each bot has its own form to fill.

Sometimes you give everything at once: “Book two seats to London on March 15.” The bot grabs all three slots in one go. Other times you only say “I want to fly to London,” and the bot has to ask follow-up questions to fill the remaining blanks — “When would you like to travel?” and “How many seats?”

The detective work of finding the details in a sentence is called entity extraction. The bot scans your words, spots things that look like dates, places, numbers, or names, and tags them. “March 15” gets tagged as a date. “London” gets tagged as a city. “Two” gets tagged as a number.

A common mix-up is thinking the bot understands your sentence like a person does. It does not read for meaning — it hunts for patterns. It knows that words after “to” are often destinations and that numbers before “seats” are quantities. Pattern hunting, not reading comprehension.

The one thing to remember: Slot filling is the chatbot’s way of extracting the important details from your message and filling in a form — so it can actually do something useful with what you said.