When users perform searches on Iconduck (and no results are found), I use a 3rd party service to see if the user misspelled their query.
If they did, I'll then suggest that they perform a search for the "suggested query".
Over time, this has allowed me to generate a map of 80,000+ (likely) misspelled queries that can be mapped to a "more likely" correct query.
I've open sourced the CSV and JSON of this map in this repo for others to use.
One way map.json can be used is:
When a user performs a search in your product, and no results are found, check
the map to see if it corresponds with another query.
If it does, you could either:
- Run the query again using the "more likely" correct query, or;
- Suggest the "more likely" correct query to them as something that can search
I've included map.csv incase that's useful for your data processing context.
I'll aim to update this map at the beginning of each annual-quarter. Specifically:
- 01 January
- 01 April
- 01 July
- 01 October
Send it over to: onassar@gmail.com
- JSON converter: https://www.convertcsv.com/csv-to-json.htm
- Query:
SELECT `query`, REPLACE(REPLACE(REPLACE(`suggestion`, '¨', ''), ',', ''), '`', '') AS suggestion FROM `searchSuggestions` WHERE `suggestion` != '' ORDER BY `query`;