ijelid-bert-base-multilingual

This model is a fine-tuned version of BERT multilingual base model (cased) on the Indonesian-Javanese-English code-mixed Twitter dataset.

Label ID and its corresponding name:

Label ID	Label Name
LABEL_0	English (EN)
LABEL_1	Indonesian (ID)
LABEL_2	Javanese (JV)
LABEL_3	Mixed Indonesian-English (MIX-ID-EN)
LABEL_4	Mixed Indonesian-Javanese (MIX-ID-JV)
LABEL_5	Mixed Javanese-English (MIX-JV-EN)
LABEL_6	Other (O)

It achieves the following results on the evaluation set:

• Loss: 0.3553
• Precision: 0.9189
• Recall: 0.9188
• F1: 0.9187
• Accuracy: 0.9451

It achieves the following results on the test set:

• Overall Precision: 0.9249
• Overall Recall: 0.9251
• Overall F1: 0.925
• Overall Accuracy: 0.951

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: 20

Training results

Training Loss	Epoch	Step	Validation Loss	Precision	Recall	F1	Accuracy
No log	1.0	386	0.2340	0.8956	0.8507	0.8715	0.9239
0.3379	2.0	772	0.2101	0.9057	0.8904	0.8962	0.9342
0.1603	3.0	1158	0.2231	0.9252	0.8896	0.9065	0.9367
0.1079	4.0	1544	0.2013	0.9272	0.8902	0.9070	0.9420
0.1079	5.0	1930	0.2179	0.9031	0.9179	0.9103	0.9425
0.0701	6.0	2316	0.2330	0.9075	0.9165	0.9114	0.9435
0.051	7.0	2702	0.2433	0.9117	0.9190	0.9150	0.9432
0.0384	8.0	3088	0.2545	0.9001	0.9167	0.9078	0.9439
0.0384	9.0	3474	0.2629	0.9164	0.9159	0.9158	0.9444
0.0293	10.0	3860	0.2881	0.9263	0.9096	0.9178	0.9427
0.022	11.0	4246	0.2882	0.9167	0.9222	0.9191	0.9450
0.0171	12.0	4632	0.3028	0.9203	0.9152	0.9177	0.9447
0.0143	13.0	5018	0.3236	0.9155	0.9167	0.9158	0.9440
0.0143	14.0	5404	0.3301	0.9237	0.9163	0.9199	0.9444
0.0109	15.0	5790	0.3290	0.9187	0.9154	0.9169	0.9442
0.0092	16.0	6176	0.3308	0.9213	0.9178	0.9194	0.9448
0.0075	17.0	6562	0.3501	0.9273	0.9142	0.9206	0.9445
0.0075	18.0	6948	0.3520	0.9200	0.9184	0.9190	0.9447
0.0062	19.0	7334	0.3524	0.9238	0.9183	0.9210	0.9458
0.0051	20.0	7720	0.3553	0.9189	0.9188	0.9187	0.9451

Framework versions

• Transformers 4.21.2
• Pytorch 1.7.1
• Datasets 2.5.1
• Tokenizers 0.12.1