saul
saul copied to clipboard
WIP: Chunker Example
Progress for the Chunker Example
Features (LBJava Reference)
- [x] PreviousTags
- [x] POSWindow (Edison-based)
- [x] Capitalization in a window (Edison-based)
- [x] WordTypeInformation (Edison-based)
- [x] Affixes (Edison-based)
- [x] Forms
- [x] Formpp
- [x] Mixed
- [x] SOPrevious
Other tasks
- [x] Basic Dataset Reader
- [x] Cleanup ChunkerApp into Train and Test functions and separate Reader
- [x] Documentation
- [x] Annotator
- [x] Interactive Mode
- [ ] ~Unit Tests (if required)~
Pending
- [ ] Performance is about 5-6 points lower than the LBJ version. Investigate feature extraction
- [ ] Deploy trained model and incorporate it.