Unstructured text to structured data

I’m looking to build a lookup service that takes unstructured user text and translates it to structured data. The data involves injuries and translating to the impacted body part.

For example, unstructured text like the following would all results in a body part value of “arm”

  • dislocated elbow
  • torn bicep
  • lateral epicondylitis

We have a database of >10k entries of injuries translating to one of 25 different body parts. This database is mostly manually maintained.

I’m curious if there’s a way to use pinecone to manage this database better. Finding matches within a proximity would help us not have to keep the database updated constantly. Thanks in advance for any feedback. Full disclosure: new to ML