Answers - Autocomplete Questions

Hi Team,

I’ve received a few questions across clients related to how our autocomplete functionality works. Find these below, will appreciate your thoughts - thanks.

  1. How are the autocomplete suggestions generated? My assumption was queries that had been searched for X number of times showed, but I think it’s a little more complex than that.

  2. Is it possible to provide a client with an export of all the current autocomplete suggestions showing in their experience? The use case is they’d review and blacklist those they don’t want showing.

  3. If a suggestion is blacklisted, can it be “unblacklisted”? So if I blacklist within querySuggestions in the config but then remove that blacklist, will the suggestion surface again?

Hey Scott -

  1. We have some more info about this in this module. You’re correct - there are a few more criteria; a query needs to be searched “more than average”, be between 2 and 75 characters, and a few other things.
  2. We can’t provide an export of exactly which queries will appear as popular queries in query suggestions, but if you export your search terms data and sort it by popularity (# of searches), this will very closely resemble the popular queries you’ll see in query suggestions.
  3. Definitely, yes! Just remove it from the popularQueriesBlacklistedTerms and it will start showing up again.
2 Likes

Thank you, this is perfect!

Hey Max or whomever is able to help,

I had another autocomplete question: Do we have any capabilities to have autofill/autocomplete take place based on words in the middle of an entity name? I know it works for the start of an entity name so if I typed in “freque” it would likely show me “frequently asked questions” but if I typed in just “questions” which is the tail end, is there a way to make sure that can trigger autofill/autocomplete?

This isn’t popular queries since that is disabled. This is just for autocomplete on entities in KG.

Hey Khari,

At the moment our query suggestions system only does prefix matching, meaning that we we only match suggestions that start with the same letters that as the user’s query string.

In the future we plan on improving this to match on any word in the query.