Update README.md

README.md

@@ -8,98 +8,87 @@ base_model:
 - distilbert/distilbert-base-uncased
 pipeline_tag: text-classification
 tags:
-- bug-detection
-- review-classification
+- Bug Detection
+- Review Classification
 - NLP
-- deep-learning
-- app-reviews
-- game-reviews
+- Deep Learning
+- App Reviews
+- Game Reviews
 - BlaikHole
 ---
-# DistilBERT Review Bug Classifier by BlaikHole
 
-**DistilBERT Review Bug Classifier** is a fine-tuned [DistilBERT](https://huggingface.co/distilbert-base-uncased) model built to automatically classify bug-related review texts. The model distinguishes between four categories:
+# 🚀 DistilBert Reviews Bug Classifier by BlaikHole
 
-- **No:** Reviews that are not bug-related.
-- **Performance:** Reviews describing performance issues.
-- **Network:** Reviews reporting connectivity or network issues.
-- **Graphics:** Reviews referring to graphical or UI/UX problems.
+<p align="center">
+  <a href="https://huggingface.co/blaikhole/distilbert-review-bug-classifier" target="_blank" rel="noopener noreferrer">
+    <img src="https://img.shields.io/badge/HuggingFace-Model-yellow?logo=huggingface" height="35">
+  </a>
+  <a href="https://huggingface.co/spaces/blaikhole/review-bug-classifier" target="_blank" rel="noopener noreferrer">
+    <img src="https://img.shields.io/badge/Demo-Space-blue?style=flat-square" height="35">
+  </a>
+</p>
 
-This model is almost perfect for any developers, QA teams, and product managers who want to quickly analyze bug reports from large data of any app store reviews.
+## 📌 Overview
 
-## Overview
+This repository provides a **fine-tuned** model trained on our **private Playstore reviews data** using **quick still efficient DistilBert architecture**. It can be used for **Reviews classification with 3 classes (with 4th class - No bug)**.
 
-In real world applications, bug reports and review feedback can be overwhelming. This classifier makes the process easy by automatically categorizing reviews into clear, actionable labels. Using DistilBERT ensures the model is both fast and efficient without affecting accuracy much.
+---
 
-- **Model Architecture:** DistilBERT for Sequence Classification
-- **Number of Labels:** 4 (No, Performance, Network, Graphics)
-- **Training:** Fine-tuned using stratified k-fold cross-validation to maximize results accuracy and prevent overfitting.
+## 🎨 Model Outputs & Labels
 
-## Why Use This Model?
+The model identifies the following labels:
 
-- **Efficient & Fast:** DistilBERT is a lighter and faster alternative to BERT with small accuracy loss and is ideal for real-time analysis.
-- **Accurate Classification:** The model is fine-tuned to effectively differentiate between major and most common bug report types.
+| Label Name | Description |
+|------------|-------------|
+| 🟥 **LABEL_0 > Graphics Issue** | Screen touch controls issue, graphics flickering, rendering issues. |
+| 🟩 **LABEL_1 > Network Issue** | Login/signup, account issues, wi-fi/data or ping problems etc. |
+| 🟦 **LABEL_2 > No Bug** | No bug discussion found. |
+| 🟨 **LABEL_3 > Performance Issue** | Overheating mobile, lag, crash, stuck game and so on. |
 
-## Quick Start
+---
 
-You can load and use the model directly using the Hugging Face Transformers pipeline. The pipeline automatically handles label mapping, so you get a human-readable output without any extra steps.
+## 🚀 Quick Usage
 
-### Example Code
+You can easily load and use this model with `transformers`:
 
+### 🔹 Named Entity Recognition (NER)
 ```python
 from transformers import pipeline
 
-# Define the repository name where your model is hosted
-model_name = "blaikhole/distilbert-review-bug-classifier"
-
-# Load the text classification pipeline
-classifier = pipeline("text-classification", model=model_name, tokenizer=model_name)
+# Label Mapping
+LABEL_MAP = {
+    "LABEL_0": "Graphics issue", 
+    "LABEL_1": "Network issue", 
+    "LABEL_2": "No Bug", 
+    "LABEL_3": "Performance issue"
+}
 
-# Example review text
-review_text = "This app has terrible performance after the latest update."
-
-# Get prediction
-result = classifier(review_text)
-print(result)
-```
+# Load Text Classification Model
+MODEL_NAME = "blaikhole/distilbert-review-bug-classifier"
+classifier = pipeline("text-classification", model=MODEL_NAME, tokenizer=MODEL_NAME)
 
-### Classify text in an excel file
-
-```python
-#pip install pandas openpyxl transformers tqdm torch
-
-import pandas as pd
-from transformers import pipeline
-from tqdm import tqdm
-import torch
-
-# use GPU if available
-device = 0 if torch.cuda.is_available() else -1
-
-# Enable progress_apply for pandas
-tqdm.pandas()
+def classify_text(text):
+    result = classifier(text)[0]
+    label = LABEL_MAP.get(result["label"], "Unknown")
+    return f"Predicted Label: {label} (Confidence: {result['score']:.2f})"
 
-# Define your model repository name
-model_name = "blaikhole/distilbert-review-bug-classifier"
+# Example Usage
+if __name__ == "__main__":
+    sample_text = "The game keeps lagging and frame rates drop frequently."
+    print(classify_text(sample_text))
 
-# Load the text classification pipeline
-classifier = pipeline("text-classification", model=model_name, tokenizer=model_name)
+```
+---
 
-# Read the Excel file with text column
-input_file = "input.xlsx"  # change this to your input file path
-df = pd.read_excel(input_file)
+## 📦 Installation
 
-# Function to classify a single review text
-def classify_text(text):
-    result = classifier(text)[0]  # Get the first result
-    return pd.Series([result['label'], result['score']])
+To use this model, install the required dependencies:
 
-# Apply the classification function with a tqdm progress bar
-df[['label', 'confidence']] = df['text'].progress_apply(classify_text)
+```bash
+pip install transformers torch
+```
+---
 
-# Save the output to an Excel file
-output_file = "output.xlsx"  # change this to your desired output file path
-df.to_excel(output_file, index=False)
+## 📜 License
 
-print(f"Results saved to {output_file}.")
-```
+MIT