🔎 ResearchReach: Intelligent Research Paper Matcher & Cold-Email Assistant

AI-powered system that analyzes resumes and recommends the most relevant research papers

📑 Table of Contents

• 🔎 ResearchReach: Intelligent Research Paper Matcher & Cold-Email Assistant
• 🌟 Overview
• 🚀 Introduction
• 🔍 How It Works
  • 1️⃣ Resume Data Extraction
• 🔍 Research Paper Matching System
  • 🛠️ Tech Stack
  • 🏆 Features
  • 🚀 Process Overview
• 📝 1️⃣ Resume Parsing and Skill Extraction
• 📜 2️⃣ Research Paper Retrieval
• 🔎 3️⃣ Convert to Sentence Embeddings
• 📈 4️⃣ Compute Cosine Similarity
• 💡 5️⃣ Final Output
• ✉️ 6️⃣ Email Generation
• 🤝 Contributors

🌟 Overview

ResearchReach is an AI-driven web platform that intelligently matches research papers with a candidate’s resume.
Using advanced natural language processing techniques such as Sentence-BERT (SBERT) embeddings and cosine similarity, the system evaluates a user’s:

• Skills
• Projects
• Technical experience
• Research interests

It then identifies the most relevant research papers from the web and automatically drafts a professional cold-email tailored to the selected paper.

This makes the process of research discovery and outreach faster, more accurate, and significantly more efficient.

---

🚀 Introduction

Finding research papers that align precisely with your skills and academic profile can be tedious.
ResearchReach fully automates this process in four steps:

• ✅ Extracts skills and project details from the resume
• ✅ Converts resume and paper text into embeddings using SBERT
• ✅ Computes semantic similarity using cosine similarity
• ✅ Recommends the most relevant research paper with a high matching score

Designed for students, researchers, and applicants seeking internships or collaboration opportunities, ResearchReach offers a streamlined, intelligent, and user-friendly experience.

---

🔍 How It Works

The matching process follows a 4-step pipeline:

1️⃣ Resume Data Extraction

• The system extracts key details from the candidate's resume, including:
  • ✅ Skills (e.g., Machine Learning, NLP)
  • ✅ Projects (e.g., Fake News Detection using BERT)

✅ Example:

Skills = ["Machine Learning", "Natural Language Processing", "Deep Learning", "Python"]  
Projects = ["Fake News Detection using BERT", "Text Summarization with LSTM"]  

This information is concatenated into a single text input:

"Machine Learning Natural Language Processing Deep Learning Python Fake News Detection using BERT Text Summarization with LSTM" 

🔍 Research Paper Matching System

🛠️ Tech Stack

Component	Tool
Frontend	React.js
Backend	Flask
Embedding Model	Sentence-BERT (all-MiniLM-L6-v2)
Paper Retrieval	Semantic Scholar API
Similarity Calculation	Cosine Similarity (Scikit-learn)
Email Generation	Gemini API
Paper Download	Unpaywall API

---

🏆 Features

✅ Fast and Efficient: Handles large datasets quickly using SBERT.
✅ Accurate Matching: High similarity scoring using cosine similarity.
✅ Automated Paper Retrieval: Uses Semantic Scholar to find relevant papers.
✅ Secure Data Handling: Ensures data privacy and integrity.
✅ Email Automation: Automatically generates internship request emails based on the matching paper.

---

🚀 Process Overview

1. Resume Parsing and Skill Extraction
2. Research Paper Retrieval
3. Convert to Sentence Embeddings
4. Compute Cosine Similarity
5. Generate and Send Email

Step 3

Step 2

Step 1

Step 4

---

📝 1️⃣ Resume Parsing and Skill Extraction

The system extracts skills and projects from the resume using pdfplumber, spaCy, and KeyBERT.

Example Skills:
Machine Learning, Natural Language Processing, Deep Learning, Python, Fake News Detection using BERT, Text Summarization with LSTM

---

📜 2️⃣ Research Paper Retrieval

The system retrieves research papers using Web Scraping with the help of beautifulsoup4 & Spacy

Example papers:

📜 Paper 1:
Title: "A Deep Learning Approach to Fake News Detection"
Abstract: "We propose a model based on BERT for detecting fake news articles. Our approach achieves state-of-the-art performance in text classification tasks."

📜 Paper 2:
Title: "Efficient Image Classification with CNNs"
Abstract: "We present an optimized CNN model for image classification. The model reduces computational cost while maintaining accuracy."

---

🔎 3️⃣ Convert to Sentence Embeddings

The system converts text into high-dimensional vector embeddings using Sentence-BERT (all-MiniLM-L6-v2):

from sentence_transformers import SentenceTransformer  

embed_model = SentenceTransformer('all-MiniLM-L6-v2')  
resume_embedding = embed_model.encode(resume_text)  
paper_1_embedding = embed_model.encode(paper_1_text)  
paper_2_embedding = embed_model.encode(paper_2_text)  

✅ Example vector embeddings:

Resume Embedding → [0.12, -0.08, ..., 0.32]  
Paper 1 Embedding → [0.11, -0.07, ..., 0.30]  
Paper 2 Embedding → [0.02, 0.45, ..., -0.12]  

---

📈 4️⃣ Compute Cosine Similarity

Cosine similarity measures how similar two vectors are:

[ \text{Cosine Similarity} = \frac{A \cdot B}{||A|| \cdot ||B||} ]

---

✅ Example calculation:

from sklearn.metrics.pairwise import cosine_similarity  

similarity_1 = cosine_similarity([resume_embedding], [paper_1_embedding])  
similarity_2 = cosine_similarity([resume_embedding], [paper_2_embedding])  

Pair	Similarity Score	Result
Resume & Paper 1	0.92	✅ High Similarity
Resume & Paper 2	0.34	❌ Low Similarity

---

💡 5️⃣ Final Output

The paper with the highest similarity score is selected as the most relevant match.

✅ Most Relevant Paper Found!
Title: "A Deep Learning Approach to Fake News Detection"
Abstract: "We propose a model based on BERT for detecting fake news articles. Our approach achieves state-of-the-art performance in text classification tasks."
Similarity Score: 0.92

---

✉️ 6️⃣ Email Generation

Once a matching paper is found, the system generates an internship request email using the Gemini API.

Template Options:

✅ Formal & Professional
✅ Technical & Research-Oriented
✅ Enthusiastic & Passionate

---

🤝 Contributors

We would like to extend our heartfelt gratitude to everyone who contributed to this project. Your hard work and dedication made this possible!

Srujan Rana 🏆 Project Lead, Backend Developer

Rudra Prasad Jena 💻 Frontend Developer & 🌐 API Integration

Abhishek Kumar 💻 Frontend Developer

🌟 Want to contribute?
We welcome contributions from the community! If you'd like to improve the project or report issues, feel free to fork the repo and submit a pull request.