Index of start and end of entity

[grungert]

Hi, an excellent tool for extraction. Is there a way for output to have a start and end index in the sentence? It would be a great improvement to highlight the inside sentence entity.

For example:
{'personal_info': [{'first_name': {text: string, start: integer, end: integer}, 'last_name': {text: 'Last name', start: int, end: int}''}]}

Thanks

[eyurtsev]

Hi @grungert , no way currently, but something I'm planning on trying to implement.

The solution I'm thinking on trying is to do a second pass with an LLM and ask the LLM to localize the extracted results from the raw text.

Also if you're interested on helping out with this feature let me know! :)

[smwitkowski]

@eyurtsev are you set on using an LLM call to get the start and end index? Instead, you may be able to do a string search, assuming the entity extracted exists verbatim.

[eyurtsev]

@smwitkowski String search could be useful as well -- it doesn't have enough flexibility in comparison to an LLM, but it's fast and basically free. If you want to work on this feature let me know!