File size: 745 Bytes
3342c07 |
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 |
---
datasets:
- universal_dependencies
language:
- tl
metrics:
- f1
pipeline_tag: token-classification
---
## Model Specification
- Model: RoBERTa Tagalog Base (Jan Christian Blaise Cruz)
- Randomized training order of languages
- Training Data:
- Combined English, Serbian, Slovenian, & Naija corpora (Top 4 Languages)
- Training Details:
- Base configurations with learning rate 5e-5
## Evaluation
- Evaluation Dataset: Universal Dependencies Tagalog Ugnayan (Testing Set)
- Tested in a zero-shot cross-lingual scenario on a Universal Dependencies Tagalog Ugnayan testing dataset (with 72.97\% Accuracy)
## POS Tags
- ADJ – ADP – ADV – CCONJ – DET – INTJ – NOUN – NUM – PART – PRON – PROPN – PUNCT – SCONJ – VERB |