Embedding and similarity between numeric and word values

Hi,
I am trying to figure out how to embed word and numerical values so that they can be used together for search in a vector database.

Example: I have a phone with 5000 mAh battery and I want this phone to be selected based on user prompt “A phone with a large battery” or similar.

I am not sure if this is even possible. Perhaps I need to create the “conversion” myself.

Thanks for any feedback. J

Going to need to have some way to associate 5000 with “large”. While 5000mAh is a reasonable phone battery is is not a large EV battery, now the model may have some training data that associates 5000 and phone with large, but if not, you need to make that connection by having perhaps added metadata, like when you see a capacity of 5000 you tag it with (large) giving you 5000(large)

Thank you for a prompt response. I thought something like this would be needed but was not sure if there are no better approaches.

1 Like