[Python] How to validate search terms when using embedding to look for objects in images

I have a search on my site that does both tradition full text search and searches using embeddings. So, for example, when you search 'red balloon' I want both the text and image results. The problem is that not all search terms make sense for object detection (like, say 'William' or even like an identifier like a driver license number) I know there are libraries that will tell me if a word is a noun but is there anything that tells me if a phrase is searchable. So like this:

• Red Apple YES
• Big Idea No
• Driver's License YES
• Suspended License No

Continue reading...