Need some ideas on designing a similar search

Hello everyone. I know that there are a ton of threads here about varying issues with searches in our apps. I think I have read all of these threads and am really beating my head on this one, though.

In a production app we use for one of our businesses in-house, I have employees that are creating items for an online store. One of the biggest issues we have is people creating duplicates of existing items. Often, it’s because the title of the item may differ by just a few words.

I need a solid search function to detect items that have similar keywords in the title, and push those matches to the top of the list. I have a large database, though. Over 30,000 items and growing every day.

I have tried:
*Fuzzy Search plugin (crashed the app)
*Search and Autocorrect plugin (no results and then crashed the app)
*Do a Search for [Item] → Title Contains [Input’s value]
-Have also done many variations of truncated values
-Have done multiple merged searches of truncated and full values
*Have use a search box and input field to try to display items that have similar titles
*Have used conditions on a search box to populate the default value with the closest matching item by going through and truncating the input’s value one character a time. (desperation)
*Have used the method put forth in search thread to populate a group with the searchable text, and extract with regex to produce results in a repeating group (crashed the app).
*Have tied in Google Site Search to search the existing items on the company’s website based on the value of the input. Not reliable in results.
-Other such solutions that proved not to be solutions.

At this point I’m not sure what else to try. I just need a good way to search for items with similar titles.

For example:
Input’s value = “Alfani AlfaTech Mens Plaid Button-Down Dress Shirt”
This should also be able to find items that are similar, such as:
Alfani Mens Plaid Button Down Shirt
Alfani Mens Plaid Logan Button-Up Shirt
Alfani AlfaTech Mens Solid Dress Shirt

I would more than welcome ideas at this point. For the love of my sanity. You don’t know what this is doing to my life.

Interesting, curious to see what the solution ends up being.

A complex word search is beyond my level – my gut would be to approach this by using a searchbox, which dynamically pulls up and then narrows the results list as you enter each word. And then do some heavy training of the staff on how to search and avoid dupes. So with your example, staff should know to first type in “Alfani” and “shirt” and that will capture all the existing similar entries. If they need to narrow further, then they can type in men, and so on. I think no matter how fancy your search function is, someone could still screw it up if they didn’t know the basics of how to search or care about avoiding a dupe.

Perhaps this approach by @TipLister could be useful:

That is an interesting idea that I hadn’t seen. However, it would be a pretty intensive amount of work to create an index out of over 30,000 items that have thousands of potential keywords. Thank you for the idea. I’ll certainly consider if it can be done.