AHR-Search: Adaptive Hybrid Retrieval Engine for Academic Papers 🔍

AHR-Search is a high-performance, intelligent search engine designed to bridge the gap between keyword-based search and semantic understanding in academic repositories. By integrating BM25 (Lexical) and Transformer-based (Semantic) retrieval, it provides a dynamic weighting mechanism to improve search precision.

---

🚀 Key Features

• Live ArXiv Integration: Fetches the latest research papers directly from Cornell University's repository.
• Adaptive Hybrid Scoring: Dynamically balances lexical and semantic weights ($\alpha$-adaptive) based on query information density.
• Transformers-powered: Uses all-MiniLM-L6-v2 for deep semantic context understanding.
• Statistical Normalization: Implements RobustScaler to ensure score alignment between disparate retrieval models.

---

🧬 Mathematical Core

The core innovation of this project is the Adaptive Alpha Fusion. The final relevance score $S$ is calculated as:

$$Score_{Final} = \alpha_q \cdot \hat{S}_{lex} + (1 - \alpha_q) \cdot \hat{S}_{sem}$$

Where $\alpha_q$ is a sigmoid-gated function of the query's Average Inverse Document Frequency (AIDF), ensuring that rare technical terms are prioritized via lexical matching while general concepts are handled semantically.

---

🛠️ Installation & Setup

1. Clone the repository:

   git clone [https://github.com/YOUR_USERNAME/AHR-Search.git](https://github.com/YOUR_USERNAME/AHR-Search.git)
   cd AHR-Search