Citation Generator
Project Overview
Supporting scientific writing with credible paper citations is crucial, particularly in key sections like the introduction, literature review, and discussions. The AI citation generator produces a list of relevant paper citations from top-tier journals and conferences in the relevant research field, based the provided academic writing.
Challenge
The key target of this project was to automate the citation process in research. A major challenge was ensuring that citations came from top-tier conferences and journals in the relevant field to guarantee their validity and legitimacy. Additionally, the cited papers had to be directly relevant to the provided research text, avoiding hallucinations from the LLM.
Methodology
- Workflow Design: Develop a prototype of the AI agent by dividing the main objective into smaller tasks.
- Prompt customization: Design the system prompts and tailor them into the defined workflow.
- Evaluation and Improvement: Validate the model performance on new academic writings.
- Deployment: Deploy the AI citation generator application using cloud services.
Tools
The project mainly relied on the GPT-4o LLM, using the OpenAI API in Python for workflow design and development. Git and GitHub were employed for version control and collaboration. Testing and deployment were supported by continuous integration/continuous deployment (CI/CD) practices. The Google Cloud Platform (GCP) was used for cloud deployment.
Results
The final product was a citation generator that automated the citation process, enhancing efficiency in academic writing. This tool relieved researchers of mundane citation tasks, enabling them to focus more time and energy on innovation in their respective research fields.