.

.gitignore

@@ -0,0 +1,4 @@
+.DS_Store
+dataset.png
+loss_value.png
+training_accuracy.png

README.md

@@ -1,3 +1,52 @@
 ---
 license: apache-2.0
 ---
+
+# 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.
+![Graph](/Users/deepakrana/Desktop/NepaliSentimentAnalysis/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.
+
+![Graph](/Users/deepakrana/Desktop/NepaliSentimentAnalysis/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.
+
+![Graph](/Users/deepakrana/Desktop/NepaliSentimentAnalysis/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 `./model`.
+
+---
+
+**Note:** The code provided is a simplified version for demonstration purposes. Additional error handling, logging, and hyperparameter tuning can be added for production use.

config.json

@@ -0,0 +1,37 @@
+{
+  "_name_or_path": "bert-base-uncased",
+  "architectures": [
+    "BertForSequenceClassification"
+  ],
+  "attention_probs_dropout_prob": 0.1,
+  "classifier_dropout": null,
+  "gradient_checkpointing": false,
+  "hidden_act": "gelu",
+  "hidden_dropout_prob": 0.1,
+  "hidden_size": 768,
+  "id2label": {
+    "0": "LABEL_0",
+    "1": "LABEL_1",
+    "2": "LABEL_2"
+  },
+  "initializer_range": 0.02,
+  "intermediate_size": 3072,
+  "label2id": {
+    "LABEL_0": 0,
+    "LABEL_1": 1,
+    "LABEL_2": 2
+  },
+  "layer_norm_eps": 1e-12,
+  "max_position_embeddings": 512,
+  "model_type": "bert",
+  "num_attention_heads": 12,
+  "num_hidden_layers": 12,
+  "pad_token_id": 0,
+  "position_embedding_type": "absolute",
+  "problem_type": "single_label_classification",
+  "torch_dtype": "float32",
+  "transformers_version": "4.30.1",
+  "type_vocab_size": 2,
+  "use_cache": true,
+  "vocab_size": 30522
+}

pytorch_model.bin

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:00de4aecbec3d003d7d94334a955fa58f25e68ecc2910bd30b98dd9c79c85172
+size 438010997

special_tokens_map.json

@@ -0,0 +1,7 @@
+{
+  "cls_token": "[CLS]",
+  "mask_token": "[MASK]",
+  "pad_token": "[PAD]",
+  "sep_token": "[SEP]",
+  "unk_token": "[UNK]"
+}

tokenizer_config.json

@@ -0,0 +1,15 @@
+{
+  "clean_up_tokenization_spaces": true,
+  "cls_token": "[CLS]",
+  "do_basic_tokenize": true,
+  "do_lower_case": true,
+  "mask_token": "[MASK]",
+  "model_max_length": 512,
+  "never_split": null,
+  "pad_token": "[PAD]",
+  "sep_token": "[SEP]",
+  "strip_accents": null,
+  "tokenize_chinese_chars": true,
+  "tokenizer_class": "BertTokenizer",
+  "unk_token": "[UNK]"
+}