PAPERGEN
AI Research Paper Generator with IEEE Format Export
The Problem Statement
“Writing a structured IEEE-format research paper from scratch requires navigating academic citation databases, synthesising literature, and formatting output to exact typographic standards — a multi-hour process that most researchers front-load before the actual writing. PAPERGEN compresses this entire pipeline into a single tool: it retrieves relevant papers from Semantic Scholar, builds a RAG context from real literature, generates a complete paper structure using a locally-run LLaMA 3.1 model via Ollama, and exports the result as a properly formatted IEEE PDF or DOCX. It stands as the practical middle ground between a blank research template and a full academic writing assistant.”
The Architecture Layout
PAPERGEN is a Python Flask web application running on port 8080. The literature survey module queries the Semantic Scholar API with the user's topic, retrieving paper titles, abstracts, and citation metadata. Retrieved abstracts are injected as RAG context into structured prompts sent to a locally-running Ollama LLaMA 3.1:8b instance. The generation pipeline produces each paper section (abstract, introduction, related work, methodology, results, conclusion) sequentially, maintaining section coherence through prompt chaining. Matplotlib generates dynamic figures — charts, tables, word clouds — from the content. The DOCX export module uses python-docx to assemble the paper with IEEE-compliant styles, and a PDF rendering step produces the final output. An OCR module (Tesseract or pytesseract) enables text extraction from user-uploaded reference images.
Architecture Design Diagram
