bert-model-intent-classification

This model is a fine-tuned version of bert-base-uncased on the None dataset.

Model description

We have finetuned Base Bert model for text classification task. We used intent-detection dataset for traning our model.

Intended uses & limitations

More information needed

How to use

Use below code to test the model

new_model = AutoModelForSequenceClassification.from_pretrained("ArunAIML/bert-model-intent-classification", num_labels=21, id2label=id_to_label, label2id=label_to_id)

tokenizer = AutoTokenizer.from_pretrained("bert-base-uncased")

Training and evaluation data

More information needed

Training procedure

Training hyperparameters

The following hyperparameters were used during training:

• learning_rate: 5e-05
• train_batch_size: 8
• eval_batch_size: 8
• seed: 42
• optimizer: Use OptimizerNames.ADAMW_TORCH with betas=(0.9,0.999) and epsilon=1e-08 and optimizer_args=No additional optimizer arguments
• lr_scheduler_type: linear
• num_epochs: 10

Training results

Label Maps used

id_to_labels = {0: '100_NIGHT_TRIAL_OFFER', 1: 'ABOUT_SOF_MATTRESS', 2: 'CANCEL_ORDER', 3: 'CHECK_PINCODE', 4: 'COD', 5: 'COMPARISON', 6: 'DELAY_IN_DELIVERY', 7: 'DISTRIBUTORS', 8: 'EMI', 9: 'ERGO_FEATURES', 10: 'LEAD_GEN', 11: 'MATTRESS_COST', 12: 'OFFERS', 13: 'ORDER_STATUS', 14: 'ORTHO_FEATURES', 15: 'PILLOWS', 16: 'PRODUCT_VARIANTS', 17: 'RETURN_EXCHANGE', 18: 'SIZE_CUSTOMIZATION', 19: 'WARRANTY', 20: 'WHAT_SIZE_TO_ORDER'} labels_to_id = {'100_NIGHT_TRIAL_OFFER': 0, 'ABOUT_SOF_MATTRESS': 1, 'CANCEL_ORDER': 2, 'CHECK_PINCODE': 3, 'COD': 4, 'COMPARISON': 5, 'DELAY_IN_DELIVERY': 6, 'DISTRIBUTORS': 7, 'EMI': 8, 'ERGO_FEATURES': 9, 'LEAD_GEN': 10, 'MATTRESS_COST': 11, 'OFFERS': 12, 'ORDER_STATUS': 13, 'ORTHO_FEATURES': 14, 'PILLOWS': 15, 'PRODUCT_VARIANTS': 16, 'RETURN_EXCHANGE': 17, 'SIZE_CUSTOMIZATION': 18, 'WARRANTY': 19, 'WHAT_SIZE_TO_ORDER': 20}

Please use above labels to reproduce results

Framework versions

• Transformers 4.52.4
• Pytorch 2.6.0+cu124
• Datasets 2.14.4
• Tokenizers 0.21.1

Results

The model was evaluated on a validation set. Below is the detailed classification report in a tabular format:

Label	Precision	Recall	F1-Score	Support
100_NIGHT_TRIAL_OFFER	1.00	1.00	1.00	4
ABOUT_SOF_MATTRESS	1.00	1.00	1.00	2
CANCEL_ORDER	1.00	1.00	1.00	2
CHECK_PINCODE	1.00	1.00	1.00	2
COD	1.00	1.00	1.00	2
COMPARISON	0.33	0.50	0.40	2
DELAY_IN_DELIVERY	1.00	1.00	1.00	2
DISTRIBUTORS	1.00	1.00	1.00	7
EMI	0.89	1.00	0.94	8
ERGO_FEATURES	1.00	1.00	1.00	2
LEAD_GEN	1.00	1.00	1.00	4
MATTRESS_COST	1.00	0.80	0.89	5
OFFERS	1.00	1.00	1.00	2
ORDER_STATUS	1.00	0.75	0.86	4
ORTHO_FEATURES	1.00	1.00	1.00	4
PILLOWS	1.00	1.00	1.00	2
PRODUCT_VARIANTS	0.50	0.50	0.50	4
RETURN_EXCHANGE	1.00	0.67	0.80	3
SIZE_CUSTOMIZATION	0.50	0.50	0.50	2
WARRANTY	0.67	1.00	0.80	2
WHAT_SIZE_TO_ORDER	0.80	1.00	0.89	4
Accuracy			0.89	66
Macro Avg	0.90	0.89	0.89	66
Weighted Avg	0.91	0.89	0.90	66