Have trouble to understand alignment_mode arguments

[innocent-charles]

The alignment_mode argument is used to match entities as returned by the LLM to the tokens from the original Doc - specifically it’s used as argument in the call to [doc.char_span()](https://spacy.io/api/doc#char_span). The "strict" mode will only keep spans that strictly adhere to the given token boundaries. "contract" will only keep those tokens that are fully within the given range, e.g. reducing "New Y" to "New". Finally, "expand" will expand the span to the next token boundaries, e.g. expanding "New Y" out to "New York".

I have document as "Innocent Charles is a Tanzanian" and i wanted to extract the country from it.

LLMs give me country : Tanzania when i check the logs but, i dont get anything in the docs.ents output.

Why and How that ?

[innocent-charles]

even for the "Innocent Charles is a Tanzanian, born on 22th, October 1999" . when i want to extract the day innocent was born, the LLMs give 22 when I check the logs, but the docs.ents do not provide anything.

[rmitsch]

Hi @innocent-charles! Please

• post your config and your code and
• run this with save_io on and post your output.

I'm not sure whether alignment_mode is relevant to this at all.

[innocent-charles]

Okay, but how to run it with save_io ?

[rmitsch]

You can pass save_io in to the LLM component in the config. See the docs. I. e. in your config:

[components.llm]
factory = "llm"
save_io = True

[innocent-charles]

Okay ! Thanks for help @rmitsch

[innocent-charles]

sorry, how to access the saved LLM(prompts and responses) i just dont see the file or something..

[rmitsch]

Ah, there is a mistake in the docs here:

Whether to save LLM I/O (prompts and responses) in the Doc._.llm_io custom attribute.

Doc._.llm_io should actually be Doc.user_data["llm_io"]. We'll fix the docs.