Skip to content

GUIDE.md

Latest commit

 

History

History
436 lines (311 loc) · 10.8 KB

GUIDE.md

File metadata and controls

436 lines (311 loc) · 10.8 KB
Raw

Complete Guide - Sentiment Analysis Tool

Detailed guide for installation, usage, troubleshooting, and FAQs.

Table of Contents

  1. Installation & Setup
  2. How to Use
  3. Common Questions
  4. Troubleshooting
  5. Features Explained

Installation & Setup

Option 1: Live Demo (Recommended - No Installation)

Simply visit: https://arungurajapu.github.io/Sentiment_Analysis

No installation, registration, or dependencies needed!

Option 2: Clone & Run Locally

1️⃣ Clone the Repository

git clone https://github.com/arungurajapu/Sentiment_Analysis.git
cd Sentiment_Analysis

2️⃣ No Dependencies Required!

This is a pure frontend project. No pip install, no npm install needed.

3️⃣ Run Locally

Method A: Open directly in browser

# Windows
start index.html

# Mac
open index.html

# Linux
xdg-open index.html

Method B: Use a local server

python -m http.server 8000
# Visit http://localhost:8000 in your browser

How to Use

Text Analysis

Step 1: Open the app

Step 2: Select "Text" mode

  • Click the "Text" button at the top (default mode)

Step 3: Enter your text

  • Paste or type your text in the textarea
  • One sentence per line for best results
  • Example: 
    I love this product!
Terrible service and support.
Amazing experience overall.

Step 4: Analyze

  • Click "Analyze Sentiment" button
  • Or press Ctrl + Enter

Step 5: View results

  • See sentiment predictions (✅ POSITIVE or ❌ NEGATIVE)
  • Each result shows confidence score (0-100%)
  • Your analysis auto-saves to history

Example Output:

✅ POSITIVE (98.5%)
"I love this product!"

❌ NEGATIVE (96.2%)
"Terrible service and support."

✅ POSITIVE (94.1%)
"Amazing experience overall."

Dataset Analysis

Step 1: Open the app

Step 2: Select "Dataset" mode

  • Click the "Dataset" button

Step 3: Upload your file

  • Click "Choose File"
  • Select CSV or Excel file (.csv, .xlsx, .xls)
  • Supported file size: up to 50MB

Step 4: Specify text column

  • Enter the column name containing text
  • Must match exactly (case-sensitive)
  • Examples: "Review", "Text", "Comment", "feedback"
  • Example file structure: 
    | id | Review | Rating |
| 1  | Love it! | 5 |
| 2  | Hate it | 1 |
    → Enter column name: "Review"

Step 5: Analyze

  • Click "Analyze Sentiment"

Step 6: View results

  • See all rows analyzed with sentiments
  • First 10 results displayed (most important ones)
  • Results stored in analysis history

Example:

File: product_reviews.csv
Column: "Review"
Status: ✅ 150 rows analyzed

Results:
✅ POSITIVE (99.2%) | "Love it!"
❌ NEGATIVE (98.5%) | "Hate it"
... (more results)

Features Explained

Feature How It Works
Text Mode Analyze 1-100+ sentences instantly. One sentence per line.
Dataset Mode Upload CSV/Excel. Select text column. Auto-analyzes all rows.
Real-time Prediction Get results in <1 second per text.
Confidence Score 0-100% score. Higher = more confident prediction.
Positive/Negative Binary classification (no "neutral" category).
Analysis History All past analyses saved locally in your browser.
Responsive Design Works perfectly on phone, tablet, desktop.

Common Questions

General Questions

Q: Is my data saved on your server? A: No. Analysis history stays in your browser only. We don't send user data to any server except the sentiment prediction API.

Q: Can I use this for commercial purposes? A: Yes! Under Apache 2.0 License. You can modify, distribute, and use commercially.

Q: Can I analyze non-English text? A: Currently English only. Model trained on English data. Multilingual version coming soon.

Q: Why are results sometimes wrong? A: DistilBERT is 91% accurate. Complex sarcasm, context, and slang sometimes confuse it. This is normal for NLP models.

Q: When will login/register be available? A: User authentication features coming in future updates!

Q: Can I download results as CSV? A: Future feature! For now, take screenshots or copy results manually.

Q: How accurate is the model? A: ~91% accuracy on standard test datasets. Performs best on clear, straightforward sentiment text.

Q: What's the difference between Positive and Negative? A: The model classifies text into two categories. Neutral sentiment is classified as either slightly positive or negative.

Technical Questions

Q: How does it work internally? A:

  1. You submit text to the app
  2. Text is sent to Hugging Face API
  3. DistilBERT processes it
  4. Returns sentiment prediction (Positive/Negative) + confidence score
  5. Results displayed instantly

Q: What model is used? A: DistilBERT (distilbert-base-uncased-finetuned-sst-2-english)

  • Trained on Stanford Sentiment Treebank (SST-2)
  • ~91% accuracy
  • Sub-second predictions
  • Lightweight and fast

Q: Is there an API I can use? A: Yes! The backend uses Hugging Face Spaces API. See main README for endpoints.

Q: Can I modify the model? A: The frontend is yours to modify. For backend model changes, visit Hugging Face Spaces.

Usage Questions

Q: How many texts can I analyze at once? A: Text Mode: 100+ sentences. Dataset Mode: Limited by file size (50MB max), typically 1000s of rows.

Q: What file formats are supported? A: CSV (.csv) and Excel (.xlsx, .xls) files.

Q: How do I specify the column name for dataset analysis? A: Enter exactly as it appears in your file (case-sensitive).

  • ❌ "review" won't match "Review"
  • ✅ "Review" matches "Review"

Q: Can I analyze multiple columns? A: Not currently. Specify one text column per analysis.

Q: How long does analysis take? A: Usually <1 second per text. Depends on:

  • File size
  • Internet connection
  • API server status

Q: Where are my results stored? A: In your browser's localStorage (local storage on your computer). Not on any server.

Q: Can I clear my history? A: Yes. Future feature coming. For now, clear browser data manually.

Mobile Questions

Q: Does it work on mobile? A: Yes! Fully responsive. Works on iPhone, Android, tablets, desktops.

Q: Can I upload files on mobile? A: Yes, but file upload depends on your browser. Some mobile browsers have limitations.

Q: Is it slow on mobile? A: No, same speed as desktop. Results still <1 second.

Troubleshooting

App won't load?

Symptom: App not appearing, blank page

Solutions:

  1. Clear browser cache (Ctrl+Shift+Delete on Windows, Cmd+Shift+Delete on Mac)
  2. Disable browser extensions (especially ad blockers)
  3. Try in incognito/private mode
  4. Check internet connection
  5. Try different browser (Chrome, Firefox, Safari)

API errors?

Symptom: Error message when clicking "Analyze"

Possible causes & solutions:

  • No internet connection → Check WiFi/data connection
  • API server down → Hugging Face Spaces might be restarting. Wait 1-2 minutes and try again.
  • CORS error → Usually temporary. Refresh page and retry.
  • Timeout error → API took too long. Check your internet speed or try with smaller file.

Check error details:

  1. Open browser console (F12)
  2. Click "Console" tab
  3. Look for detailed error message
  4. Share error details if contacting support

Results not saving?

Symptom: Analysis history empty after refresh

Causes:

  • Private/Incognito mode enabled (disables localStorage)
  • Browser storage full
  • Browser storage disabled

Solutions:

  1. Use normal mode (not private/incognito)
  2. Clear browser storage (Settings → Privacy → Clear browsing data)
  3. Check if localStorage is enabled:
    • Open console (F12)
    • Type: localStorage.setItem('test', 'value')
    • If error appears, localStorage is disabled

Upload issues?

Symptom: Can't upload file or file won't process

Check these:

  • File format is CSV or Excel (.csv, .xlsx, .xls) ✓
  • Column name matches exactly (case-sensitive) ✓
  • File size < 50MB ✓
  • No special characters in column name ✓
  • File is valid (not corrupted) ✓

Example:

❌ "product reviews" (space in file)
❌ "Reviews" (doesn't match "review" in file)
✅ "review" (matches column name exactly)

Wrong results?

Symptom: Predictions don't match your expectations

Remember:

  • Model is 91% accurate (not 100%)
  • Sarcasm, slang, context often confuse it
  • Model trained on English data
  • Idioms might be misclassified
  • Example:
    • "This is terrible" → NEGATIVE ✓
    • "This is terribly good" → Might classify wrong (context issue)

Solutions:

  • Verify manually - sometimes model is right, you're wrong!
  • Rephrase text more clearly
  • Try shorter sentences
  • Report persistent issues on GitHub

Other Issues

Issue: Page loads slow

Solutions:

  • Check internet speed
  • Hugging Face API might be busy. Try again in few minutes.
  • Disable extensions

Issue: Text not analyzing

Solutions:

  • Ensure text is not empty
  • Try shorter text first
  • Check internet connection
  • Clear console errors

Issue: File upload button not working

Solutions:

  • Refresh page
  • Try different browser
  • Check browser permissions (Settings → Privacy)

Browser Compatibility

Browser Status Notes
Chrome ✅ Perfect Recommended
Firefox ✅ Perfect Works great
Safari ✅ Perfect Works on Mac/iOS
Edge ✅ Perfect Works great
Opera ✅ Perfect Works great
Internet Explorer ❌ Not supported Too old, use modern browser

Performance Tips

For faster analysis:

  1. Keep sentences short (2-10 words optimal)
  2. Use clear, standard English
  3. Avoid special characters
  4. Use smaller files (< 10MB for datasets)
  5. Ensure good internet connection
  6. Close other browser tabs

Limitations & Known Issues

✓ English text only ✓ Works best with 2-10 word sentences ✓ Handles sarcasm inconsistently ✓ Max 10 results displayed at once (but processes all) ✓ File size limit: 50MB ✓ No neutral sentiment category (Positive or Negative only) ✓ Context sometimes lost in short phrases

Getting Help

If you find a bug:

  1. Open browser console (F12)
  2. Screenshot the error
  3. Go to GitHub Issues
  4. Create new issue with details

For questions: 📧 Email: arungurajapu@gmail.com

Last Updated: December 31, 2025 Status: ✅ Complete & Tested