Fuzzy search and component recommendation

KNIME Analytics Platform
1

Hello everyone,

I’m currently working through various topics on fuzzy matching. Here I also came across the KNIME software and its possibilities.

I would like to present my use case and the problem to you. Maybe someone has a suitable solution for me. :blush:

Starting point:
A customer sends a excel list of required components. The manufacturer must now compare this with his excel file in order to assign the correct component.

As errors can occur with manually managed excel lists, descriptions that are similar to the components contained in the excel must also be taken into account and output as/or a recommendation.

Example:
Customer sends excel list with a component: DGSL-12-100-C-Y3A
Manufacturer excel list: DGSL-12-100-Y3A

There would be no match here. However, as there is a high similarity between the two, it would be advantageous for the manufacturer to receive the component without -C as a recommendation (Column: Possible Part) or score (90% similarity) in order to have an indication of what to look for.

Is there a possibility here at KNIME to solve exactly this problem?

I would be very grateful for any suggestions and solutions!

Best regards
Tobi

customer_list.xlsx (11.6 KB)
inventory.xlsx (9.9 KB)

2

Hey @AugsteTo

There are a few options out there leading to similar results, albeit differing in the implementation difficulty & performance implications.

No.2 is probably the easiest to implement but if you’re trying to do the operation on 10s of millions of rows, it’ll likely take a while & take a lot of space on your disk.

you’ll also need to install nodepit (Product and Node Installation Guide — NodePit) & get Palladian’s extension for option #2

Option no. 3 is probably the best from a performance perspective but is slightly trickier to implement.

Have a look at the attached workflow , hope this helps! :slight_smile:

image
image1142×394 22.2 KB

match closest string.knwf (36.6 KB)

5 Likes
3

Many thanks for your response! I’ll try it out for my use case and give you feedback. Thanks! :sunglasses: :+1:

Best regards
Tobi

1 Like