Flask-Based Research Summarization & Citation Web App
🔗 Video Link: Watch Demo
📁 Drive Link: Access Project Files
The goal of this project is to build a Flask-based web application that enhances interaction with research papers. Users can:
- Upload papers
- Receive summaries tailored to their academic level
- View citations to related research
This is achieved through advanced Natural Language Processing (NLP) techniques like TF-IDF, TextRank, and Transformer-based models (T5). The system also uses cosine similarity to recommend related work.
A sleek and modern frontend is developed using HTML, CSS, and JavaScript.
| Purpose | Technique / Structure |
|---|---|
| Extractive summarization | TF-IDF Matrix (Sparse Matrix) |
| Graph-based summarization | Adjacency Matrix (for TextRank) |
| Citation similarity | Cosine Similarity Matrix (2D Array) |
| Persistent storage | Relational Database (SQLAlchemy ORM) |
Tables:
- Paper – stores uploaded papers and their summaries
- Citation – stores similarity-based links between papers
- 📤 Upload Research Paper (plain text)
- 📚 Summarization by Academic Level
- School & PhD → TF-IDF + Cosine Similarity (Extractive)
- Undergraduate → TextRank (Graph-based)
- General → T5 (Abstractive)
- 🔍 Citation Recommendation
- Uses cosine similarity on summary vectors
- 🗄️ Database Management
- PostgreSQL via SQLAlchemy
- 💻 Modern Frontend
- HTML, CSS (dark theme), and JavaScript for interactivity
tfidf_cosine_summarizer()text_rank_summarizer()t5_abstractive_summarizer()
- Computes similarity using
cosine_similarity()on TF-IDF vectors - Stores strong matches (score > 0.5) in
Citationtable
| Route | Description |
|---|---|
/ |
Homepage |
/upload |
Upload page |
/summarize |
Summarizes uploaded paper |
/generate_citations |
Computes and stores related papers |
/citations/<id> |
Displays recommended papers for a given paper |
- Upload a research paper
- Select academic level for summarization
- View generated summary
- Explore related research via citation recommendations
- (Future Scope) Admin operations to manage/delete entries
This web application fuses NLP and web development to simplify academic research access. From tailored summarization to smart citation discovery, the app delivers powerful features in a user-friendly format.
The choice of sparse matrices, graph algorithms, and transformer models makes the system robust and adaptable.
🚧 Coming Soon:
- PDF upload support
- Cloud deployment (e.g., Render, Heroku, or AWS)
- Python (Flask, Transformers, Scikit-learn)
- HTML + CSS + JavaScript
- SQLAlchemy + PostgreSQL