PaperMatch - Research Paper Recommendation System

Introduction

Welcome to PaperMatch, a BERT-based recommendation system designed to help you discover the most relevant academic research papers based on your search queries. By leveraging BERT embeddings, PaperMatch effectively captures the semantic meaning of both user queries and research paper abstracts. This allows for efficient and accurate recommendations of research papers tailored to your interests.

PaperMatch simplifies the process of finding academic research, providing a faster and more effective way to explore relevant literature in any field.

Features

• User Query Search: Enter a query (keywords or phrases) to receive a ranked list of relevant research papers.
• BERT Embeddings: Utilizes pre-trained BERT (Bidirectional Encoder Representations from Transformers) to create vector embeddings for both user queries and research paper abstracts.
• Cosine Similarity: Compares the query embeddings with paper embeddings using cosine similarity to recommend papers most relevant to the query.
• Flask Backend: A Flask API serves the model and handles requests from the frontend.
• React Frontend: The user interface is built with React using Vite for fast, modern web development.

Tech Stack

Backend

• Flask: A lightweight Python framework for creating the recommendation API and serving recommendations.
• BERT: The NLP model used to create meaningful embeddings from text data (using Hugging Face).
• PyTorch: For running the model with CUDA acceleration.

Frontend

• React: A JavaScript library for building user interfaces.
• Vite: A build tool that provides a fast development environment.
• Axios: For making API requests from the frontend to the backend.

How It Works

1. Query Input: The user enters a search query, which can be keywords or phrases related to research topics.
2. Text Embedding: The system tokenizes the query and converts it into an embedding using the fine-tuned BERT model.
3. Similarity Calculation: The cosine similarity is calculated between the query embedding and the embeddings of research paper abstracts in the database.
4. Recommendation: The system returns a ranked list of relevant papers based on the similarity scores.

Conclusion

The Research Paper Recommendation System provides a user-friendly interface for academics and researchers to easily find relevant papers, enhancing their research experience. By utilizing advanced NLP models like BERT, this system brings state-of-the-art recommendations right to your fingertips.

---

Installation

Prerequisites

To get the PaperMatch system running, ensure that you have the following software installed:

1. Python 3.8: You can download it from python.org.
2. Node.js: Required for React and frontend tools. Download it from nodejs.org.
3. Git: For cloning the repository. Install it from git-scm.com.

Steps to Set Up the Repository

Follow these steps to get the repository up and running on your local machine:

1. Clone the Repository

Clone the repository to your local machine using the following Git command:

git clone https://github.com/DeveloperLevin/PaperMatch.git
cd FINAL YEAR PROJECT

2. Create a Virtual Enviornment

python -m venv venv
venv\Scripts\activate

3. Install all Dependencies

pip install -r requirements.txt

4. Install all package for the frontend

cd frontend
npm install

5. Build Tailwind CSS for the project

npm run build:css

6. Run the API (first) and boot up the frontend

cd ..
python run.py
cd frontend
npm run dev

---

Future Work

In the future i plan to add a traditional recommendation system in the pipeline to create a more hybrid approach and add basic auth with the ability to view the papers within the website.