---
emoji: 📊
title: 🤖 BERT + ELMo Sentence Similarity Analyzer
---
# 🤖 BERT + ELMo Sentence Similarity Analyzer

This Gradio app allows you to compute semantic similarity between sentences using **BERT** and **ELMo** embeddings. Users can add sentences, select a BERT variant, and see pairwise similarity matrices.

---

## 🚀 Features

- Supports multiple BERT variants:
  - **BERT Large Uncased** (1024 dims, recommended for similarity)
  - **BERT Large Cased** (1024 dims, preserves capitalization)
  - **BERT Base Uncased** (768 dims, faster)
  - **BERT Base Cased** (768 dims)
- **ELMo embeddings** (1024 dims, context-aware)
- Mean-pooling over tokens for BERT
- Pairwise cosine similarity using PyTorch & Scikit-learn
- Clear and reset stored sentences anytime
- Easy-to-use Gradio interface with live status updates

---

## 📖 How to Use

1. **Choose Model**: Select your preferred BERT variant (uncased recommended for semantic similarity).  
2. **Add Sentences**: Type a sentence and click **Add Sentence**.  
3. **Repeat**: Add at least 2 sentences to compute similarity.  
4. **Compute Similarity**: Click **Compute Similarity** to generate pairwise similarity matrices.  
5. **Clear All**: Reset all stored sentences and embeddings.  

> Values closer to `1.0` indicate higher semantic similarity.

---

## 🧰 Requirements

Install dependencies from `requirements.txt`:

```bash
pip install -r requirements.txt