|
--- |
|
license: apache-2.0 |
|
datasets: |
|
- universal-dependencies/universal_dependencies |
|
language: |
|
- tl |
|
pipeline_tag: token-classification |
|
library_name: spacy |
|
tags: |
|
- part-of-speech |
|
- nlp |
|
- spacy |
|
- tagger |
|
--- |
|
https://github.com/jdoerfler/ar-fa-id-tl-SpaCy-Training |
|
|
|
SpaCy morphologizer (universal POS tagger) and dependency labeler. Trained on ~12.5k samples from Universal Dependencies' tl_newscrawl-ud-train.conllu, tested on ~1.1k samples Universal Dependencies' tl_newscrawl-ud-test.conllu, tl_trg-ud-test.conllu, and tl_ugnayan-ud-test.conllu |
|
|
|
- uPOS accuracy: 0.9389 |
|
- Dep head accuracy: 0.7851 |
|
- Dep label accuracy: 0.7028 |
|
|
|
Having tokenizing issues with this one but it doesn't affect the metrics significantly, which is a red flag... |
