Sentiment Analysis with BERT: Nepali Sentiment Dataset

This repository contains code for training and evaluating a sentiment analysis model using the BERT (Bidirectional Encoder Representations from Transformers) model on the Nepali Sentiment Dataset. The model achieves an accuracy of 99.75% on the test dataset.

Dataset

The dataset used for training and testing the sentiment analysis model is a balanced dataset in CSV format. The dataset is loaded using the pandas library. The training dataset consists of 2084 balanced data, and the test dataset consists of 2001 balanced data. Label 0 = Negative, Label 1 = Positive, Label 2 = Neutral dataset.png

Model

The BERT model is used for sequence classification and is loaded from the bert-base-multilingual-uncased pre-trained model. The model is initialized with num_labels=3 since we have three sentiment classes: positive, negative, and neutral.

Preprocessing

The dataset is preprocessed using the NepaliSentimentDataset class. The class takes the texts, labels, tokenizer, and maximum sequence length as inputs. The texts are preprocessed using regular expressions to remove special characters, usernames, and extra whitespace. The tokenizer from the Hugging Face transformers library is used to tokenize the texts and convert them into input IDs and attention masks. The preprocessed data is returned as a dictionary with the input IDs, attention masks, and labels.

Training

The model is trained using the train_model function. The function takes the model, train dataloader, and test dataloader as inputs. The model is trained for 10 epochs with an early stopping mechanism. The AdamW optimizer is used with a learning rate of 2e-5 and epsilon value of 1e-8. The function also includes additional connection layers before the classification layer of the BERT model. After each epoch, the model is evaluated on the test dataset.

Training Progress and Evaluation Metrics

This section provides insights into the training progress of the sentiment analysis model and includes graphs showing the loss values and accuracy values throughout the training process.

Loss Value Graph

The graph below displays the training progress by showing the variation in the loss values across different epochs. It helps visualize the convergence of the model during training.

loss_value.png Loss Value Graph

Accuracy Value Graph

The following graph illustrates the accuracy values achieved by the model during the training process. It presents a clear picture of how the model's performance improves over time.

training_accuracy.png Accuracy Value Graph

These graphs provide a visual representation of the training progress and performance of the sentiment analysis model, allowing for better understanding and analysis of the results.

Results

After training, the trained model achieves an accuracy of 99.75% on the test dataset.

Saving the Model

The trained model and tokenizer are saved using the save_pretrained function from the Hugging Face transformers library. The model and tokenizer are saved in the directory


Downloads last month
42
Safetensors
Model size
109M params
Tensor type
I64
·
F32
·
Inference Examples
This model does not have enough activity to be deployed to Inference API (serverless) yet. Increase its social visibility and check back later, or deploy to Inference Endpoints (dedicated) instead.