bharatttt/results

Customer Support Sentiment Classifier

This model is a fine-tuned version of distilbert-base-uncased trained on a custom customer support dataset containing customer feedback, complaints, and support-related messages.

The model classifies customer messages into one of two sentiment categories:

• Negative (0)
• Positive (1)

Model Description

• Base Model: distilbert-base-uncased
• Task: Sentiment Analysis
• Framework: Hugging Face Transformers
• Language: English
• Training Samples: 600
• Fine-Tuning Method: Supervised Fine-Tuning

Intended Use

This model can be used for:

• Customer feedback classification
• Customer review analysis
• Support ticket prioritization
• Sentiment analysis of customer messages
• Customer experience monitoring

Example

Negative Example

Input:

The app keeps crashing after login.

Output:

Negative

Positive Example

Input:

Thank you for the amazing support team.

Output:

Positive

Limitations

• Trained on a relatively small dataset (600 examples).
• Performance may decrease on domains outside customer support.
• Only supports English text.
• Predictions depend on the quality and balance of the training data.

Training and Evaluation Data

The model was trained on a custom JSONL dataset of customer support messages labeled as either Positive or Negative sentiment.

Dataset split:

• Training Set: 80%
• Test Set: 20%

Training Procedure

Hyperparameters

• Learning Rate: 2e-5
• Batch Size: 16
• Epochs: 3
• Weight Decay: 0.01
• Optimizer: AdamW
• Scheduler: Linear

Training Results

Epoch	Validation Loss
1	0.2164
2	0.0743
3	0.0571

Final Validation Loss: 0.0571

Framework Versions

• Transformers: 5.10.1
• PyTorch: 2.11.0+cu128
• Datasets: 4.0.0
• Tokenizers: 0.22.2