The client

Backstory

 You are an insurance executive who hit the mid-life crisis and left a big corporation. Now you're bored and want to set up a start-up. Your bright brand new idea of a startup is to automate the business insurance quote process. In order to give a quote for business insurance quote, the insurer needs to know what type of business it is. It determines the risk of going bankrupt. For example, cake shops go out of business more often than the coffee shops.

You know that AI can do anything so once it is automated, so your idea seems like an easy enough task, right?

Context

Start a timer for 3 minutes

Explain the problem to NLP Engineer and annotator, encourage them to ask questions.

You want to collect:

• type of business (cake shop, a cafe, a deli, a bakery
• address
• number of employees
• owner name based on this data.

IMPORTANT!

When the time runs out

continue to next slide...

What is in the data?

Start a timer for 5 minutes

"Hmm... Looks like some of the data I want is not included in the documents I gave you..."

Click ahead

Negotiation

NLP engineer will try to convince you to descope business owner, but you know the data better.

"I'm the expert, I know business owner name information is rare, but it exists in some of the documents I provided. Let's keep this field."

Agree that there is no number of employees in data and drop it.

Game master will tell you when to move on

What is a cake shop, a cafe, a deli, a bakery?

Can you come up with clear annotation rules that are consistent and make you happy?

Negotiation

NLP engineer will provide some rules to distinguish between different types of businesses.

They are a smart professional, so you agree with everything they say. Each rule they propose you accept as a good way to tell if a business is a coffee shop or not.

“Now I see how difficult it is, I am happy to keep paying you to work on this!

There is no out of the box magic”

Click ahead

Annotation UI

When they finish presenting the new tool go to the next slide

Listen to what the NLP engineer has to propose.

They are presenting you a new way to collect date.

Start annotating together with the annotator and NLP engineer using the tool provided

After 10 minutes click ahead

Briefing time

Ask NLP engineer what did we achieve

Part 2 over!