krickert commented on PR #1182:
URL: https://github.com/apache/opennlp/pull/1182#issuecomment-5032168513

   This was proposed about a week ago.   Discussions pointed to this shape - I 
feel like it's a great direction as a lot of research went into landing this 
shape:
   
   - A single shared container for full pipelines: sentence, token, POS, lemma, 
and name annotations live in one `Document` with typed, offset-anchored layers 
instead of parallel arrays passed hand to hand.
   - Exact character-offset provenance: every annotation points back into the 
original  text, the prerequisite for offset-aware normalization and span-level 
traceability.
   - A uniform `DocumentAnnotator` seam: new analysis steps (gazetteer joins, 
geocoding,  quality scoring, PII detection, relation extraction) plug in as 
annotators over the  same container rather than each inventing its own I/O 
shape.
   - Cross-component evidence sharing: one component's layer (for example emoji 
or entity annotations) is directly readable as features by another, with no 
glue code.
   - Cleaner downstream integration: document pipelines that use OpenNLP can 
consume one typed annotation record per document instead of adapting several 
tool-specific outputs.
   - Unblocks the staged follow-up branches that currently carry private copies 
of this container; once merged they rebase to plain, small diffs.


-- 
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: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

Reply via email to