Embeddings & Semantic Search

Live

Initech's support team wants to find past tickets like a new one, however the customer words it. Your call: search by matching words, by meaning, or both?

Key idea: An embedding turns text into numbers so that similar meanings land close together. That's how search finds the right ticket even when no words match.

Things to try

  1. Search "we got billed two times" with Both. Where did keyword search go wrong, and why?
  2. Search "back up everything before we leave". Which words fooled keyword search?
  3. Hover the dots on the map. Which ticket sits away from its own group, and does its spot make sense?
  4. Try a query that shares no words with any ticket, like "handset software quits randomly" or one of your own. Does meaning search still find the right one?
  5. Search "401 error". When is plain keyword search good enough?
Try a prompt
Enter to send · Shift+Enter for a new line

Change how it works, then send again

Querywhat you typedEmbedding modeltext → numberspplx-embed-v1-0.6bQuery vectora point in meaningTickets25 support ticketsTicket vectorsembedded onceNearest by meaningcosine similarityKeyword matchshared wordsResultstop matches
Send a message to watch it run

Map of meaning · close together = similar meaning

25 ticketsStep inside
LoginBillingData exportIntegrationsMobile appYour query
Search to drop your query on the map

Behind the scenes

Send a message and every step the system takes will show up here.