📊 Review Sentiment Analyzer

📌 Overview

Review Sentiment Analyzer is a comprehensive NLP Dashboard designed to analyze customer feedback, social media comments, and product reviews. Unlike simple classifiers, this tool supports Bilingual Analysis (English & Indonesian) using state-of-the-art RoBERTa Transformer models.

The application allows for both real-time text analysis and bulk processing of datasets (CSV/Excel), providing deep insights through N-Gram extraction and text complexity statistics.

✨ Key Features

🌍 Bilingual Support

• English Analysis: Powered by cardiffnlp/twitter-roberta-base-sentiment-latest.
• Indonesian Analysis: Powered by w11wo/indonesian-roberta-base-sentiment-classifier.

📂 Batch Data Processing

• Multi-Format Upload: Supports .csv and .xlsx (Excel) files.
• Automated Labeling: Classifies thousands of rows in seconds.
• Export Ready: Download the analyzed results directly as CSV.

📈 Deep Analytics & Visualization

• Sentiment Distribution: Interactive Pie Charts using Plotly.
• Keyword Extraction (N-Grams): Discover the most frequent words or phrases (Unigrams, Bigrams, Trigrams) associated with specific sentiments (e.g., "What words appear most in Negative reviews?").
• Text Complexity Stats: Automatically calculates and visualizes:
  • Average Sentence Length.
  • Average Word Count per sentiment.

💬 Real-Time Playground

• A chat-like interface to test the model with single sentences on the fly.

🛠️ Tech Stack

• Framework: Streamlit
• Models: Hugging Face Transformers (PyTorch backend)
• Data Handling: Pandas, OpenPyXL
• Visualization: Plotly Express
• NLP Utilities: Scikit-learn (CountVectorizer), NLTK, RegEx

🔍 Analytical Capabilities

Beyond simple classification, this dashboard provides deep linguistic insights:

• Automated Sentiment Labeling: Utilizes Transformer-based models (RoBERTa) to classify text into Positive, Negative, or Neutral with high contextual accuracy.
• N-Gram Keyword Extraction: Identifies top frequent words (Unigrams) and phrases (Bigrams/Trigrams) to uncover the root causes of specific sentiments (e.g., detecting "slow shipping" appearing frequently in negative reviews).
• Text Complexity & Correlation Analysis: Aggregates linguistic metrics by sentiment group to reveal behavioral patterns:
  • Sentiment vs. Sentence Length: Analyzes if specific sentiments (e.g., Negative) tend to have significantly longer sentences, indicating detailed complaints vs. short praises.
  • Sentiment vs. Word Count: Compares the average verbosity across different sentiment categories.
• Interactive Distribution Metrics: Visualizes the overall sentiment balance using dynamic Pie Charts and Bar Graphs to spot trends instantly.

🚀 Installation & Usage

1. Clone the Repository

   git clone https://github.com/viochris/Simple-Sentiment-Analysis.git
   cd Simple-Sentiment-Analysis

2. Install Dependencies

   pip install -r requirements.txt

3. Run the Dashboard

   streamlit run app.py

4. How to Use

   • Select your target language (English/Indonesia) from the dropdown.
   • Tab "File": Upload your dataset to see charts, N-grams, and download results.
   • Tab "Text": Type manually to check sentiment in real-time.

🧪 Sample Data for Testing

Don't have a dataset ready? You can use the sample files provided in this repository to test the dashboard immediately:

• English Data: sample_data/english_reviews.csv (CSV format)
• Indonesian Data: sample_data/indo_reviews.xlsx (Excel format)

Tip: Download these files and upload them in the "File" tab to see the analytics in action!

📷 Screenshots

📊 Batch Data Analysis

Upload CSV or Excel files to get instant visual analytics and sentiment distribution:

💬 Real-Time Text Analysis

Test the model with individual sentences using the interactive text mode:

---

Author: Silvio Christian, Joe "Turning unstructured text into actionable insights."