krickert opened a new pull request, #1177:
URL: https://github.com/apache/opennlp/pull/1177

   This change adds an emoji annotation layer that classic components can opt 
into, built in three strictly separated tiers so no third-party data ever 
enters the project:
   
   - **Bundled facts**: a project-authored attribute-row table 
(`codepoints;attribute;value;source;notes`) covering 40 symbols with four 
attributes each, loaded by `EmojiAnnotations` with per-value provenance and an 
audit test. The file carries no third-party data; notes cite UCD character 
names.
   - **Derived facts**: `EmojiFlags` decodes ISO 3166 codes from flag emoji 
sequences (regional-indicator pairs and tag sequences) as a pure algorithm over 
the code points, with no data file at all.
   - **Joined facts**: `EmojiAnnotationJoin` resolves ISO codes against a 
user-installed gazetteer at load or run time. Foreign identifiers are never 
baked into the bundled files, so dataset version churn cannot stale the 
distribution.
   
   The annotation surface is `EmojiAnnotator`, a parallel accessor beside 
`Term`, deliberately not a set of `Dimension` constants (rationale in the 
javadoc). Wiring is opt-in only: a feature-generator XML factory for 
NameFinder, a `DoccatFactory` array entry, and the `SentimentFactory` seam. 
Default behavior of every component is unchanged; nothing activates without 
explicit configuration.
   
   Out of scope for this iteration, tracked on the ticket: manual section, 
skin-tone/ZWJ sequence reduction, and a loader for user-downloaded sentiment 
rankings (the project ships no sentiment data; the bundled table's coarse 
sentiment column is project-authored).
   
   Verification: the full opennlp-runtime suite passes on current main (1449 
tests), including the 39 emoji-specific tests (record store audit, flag 
decoding, annotator surface, and the opt-in wiring for all three components).
   


-- 
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