alessandrobenedetti commented on PR #12314: URL: https://github.com/apache/lucene/pull/12314#issuecomment-1563074046
> Would be nice to get yours and other's ideas on them: > > 1. Is the main usage for breaking a long text into several paragraphs? Or is it also to search across several different fields (e.g. [embedding_of_title, embedding_of_content]) where we create a single graph for several fields? I would say the first, I have seen examples of that, especially when the large language model chosen by the customer has a limit in input tokens smaller than the document length for the customer. A single graph over several fields would be the equivalent of a catch-all field for the lexical search, it was not my primary focus but should be doable once we have multi-valued fields. > 2. Would it be possible to retrieve which vector was the closest match? For example, if we break a long text into paragraphs and want to highlight which paragraph was the closest match. This could be crucial for some use cases. I agree, that can be a nice addition to the explain! I guess attaching metadata to vectors is a different story, but I agree it could be a good idea! -- This is an automated message from the Apache Git Service. To respond to the message, please log on to GitHub and use the URL above to go to the specific comment. To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org --------------------------------------------------------------------- To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org For additional commands, e-mail: issues-h...@lucene.apache.org
