AI-lab-2359/china-africa

# 🌍 Dual-Tower Deep Network for China-Africa Vocational Education Matching 基于双塔深度网络与RAG增强的中非职教供需精准匹配算法 Features • Architecture • Installation • Quick Start • API Reference • Citation --- ## 📋 Overview This project implements a **Dual-Tower Deep Neural Network** enhanced with **Retrieval-Augmented Generation (RAG)** for precise matching between African labor market demands and Chinese vocational education supply. The system leverages multilingual understanding (XLM-RoBERTa) to bridge language barriers across English, French, Portuguese, Arabic, Swahili, and Chinese. ### Key Contributions - **Dual-Tower Architecture**: Separate encoders for demand and supply with cross-attention interaction - **RAG Enhancement**: Knowledge-augmented reasoning for context-aware matching - **Skill Gap Analysis**: Automated identification of training gaps with course recommendations - **Multilingual Support**: XLM-RoBERTa backbone for cross-lingual semantic understanding --- ## ✨ Features | Feature | Description | |---------|-------------| | 🏗️ **Dual-Tower Network** | Independent encoding of job requirements and course offerings | | 🔄 **Cross-Attention** | Deep semantic interaction between demand and supply vectors | | 📚 **RAG Integration** | Vector database + Knowledge graph for retrieval augmentation | | 🎯 **Gap Analysis** | Skill coverage analysis with targeted recommendations | | 🌐 **Multilingual** | Support for 6+ languages via XLM-RoBERTa | | ⚡ **Efficient Inference** | Optimized batch processing and vector caching | --- ## 🏛️ Architecture ``` ┌─────────────────────────────────────────────────────────────────┐ │ DECISION OUTPUT LAYER │ │ ┌──────────────┐ ┌──────────────┐ ┌──────────────────────┐ │ │ │ Match Score │ │ Skill Gap │ │ Recommendations │ │ │ │ [0-1] │ │ Analysis │ │ Course Matching │ │ │ └──────────────┘ └──────────────┘ └──────────────────────┘ │ └────────────────── …