scostiniano/bert-tagalog-base-uncased-ner-v1

Description

• The dataset consists of 148 Filipino storytelling books, 5,005 total sentences, 45,792 total tokens, and 5,646 unique tokens.
• This NER model only supports the Filipino language and does not include proper nouns, verbs, adjectives, and adverbs as of the moment
• The input must undergo preprocessing. Soon I will upload the code to GitHub for preprocessing the input
• To replicate the preprocessed input use this example as a guide
• Input: "May umaapoy na bahay "
• Preprocessed Input: "apoy bahay"

bert-tagalog-base-uncased-ner-v1

This model is a fine-tuned version of jcblaise/bert-tagalog-base-uncased on the None dataset. It achieves the following results on the evaluation set:

• Loss: 0.2824
• Precision: 0.9091
• Recall: 0.8988
• F1: 0.9039
• Accuracy: 0.9488

Model description

More information needed

Intended uses & limitations

More information needed

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 2e-05
• train_batch_size: 16
• eval_batch_size: 16
• seed: 42
• optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
• lr_scheduler_type: linear
• num_epochs: 10

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1	Accuracy
No log	1.0	205	0.5311	0.6465	0.5458	0.5919	0.8387
No log	2.0	410	0.3052	0.7736	0.7811	0.7774	0.9110
0.4693	3.0	615	0.2531	0.8493	0.8363	0.8427	0.9319
0.4693	4.0	820	0.2384	0.8755	0.8715	0.8735	0.9402
0.064	5.0	1025	0.2671	0.8909	0.8823	0.8866	0.9435
0.064	6.0	1230	0.2527	0.8864	0.8920	0.8892	0.9459
0.064	7.0	1435	0.2708	0.9088	0.9011	0.9049	0.9491
0.0111	8.0	1640	0.2733	0.8992	0.8977	0.8984	0.9490
0.0111	9.0	1845	0.2765	0.8991	0.8965	0.8978	0.9485
0.0037	10.0	2050	0.2824	0.9091	0.8988	0.9039	0.9488

Framework versions

• Transformers 4.24.0
• Pytorch 1.12.1+cu113
• Datasets 2.7.1
• Tokenizers 0.13.2