roberta-tagalog-profanity-classifier

This model is a fine-tuned version of jcblaise/roberta-tagalog-base on mginoben/tagalog-profanity-dataset dataset. It achieves the following results on the evaluation set:

• Loss: 0.3019
• Accuracy: 0.8898
• Precision: 0.8523
• Recall: 0.8944
• F1: 0.8728

Model description

The Model classifies tagalog texts that contains profanities as either Abusive or Non-Abusive.

It only classifies texts with the following profanities:

• bobo
• bwiset
• gago
• kupal
• pakshet
• pakyu
• pucha
• punyeta
• puta
• putangina
• tanga
• tangina
• tarantado
• ulol

Intended uses & limitations

For content moderation accross different social medias

Training and evaluation data

• Training: 11,110
• Validation: 2,778

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 1e-05
• train_batch_size: 64
• eval_batch_size: 64
• 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	Accuracy	Precision	Recall	F1
No log	1.0	174	0.3006	0.8776	0.8620	0.8458	0.8538
No log	2.0	348	0.2899	0.8834	0.8801	0.8382	0.8586
0.2993	3.0	522	0.2869	0.8873	0.8491	0.8918	0.8700
0.2993	4.0	696	0.3019	0.8898	0.8523	0.8944	0.8728

Framework versions

• Transformers 4.28.0
• Pytorch 2.0.1+cu118
• Datasets 2.12.0
• Tokenizers 0.13.3