This repository contains a simple pure Python script for checking academic references. It can be used to
- Verify if details are (generally) correct
- The reference exists or is probably hallucinated
- Create a JSON file with the references
- Run the
main.pyscript and pass the JSON file
The JSON file should follow the Semantic Scholar API style.
A simple way to create them is use an LLM, especially when parsing the references from a screenshot. A suggestion for a prompt is given in the prompt-reference-parser.md file.
The typical workflow would be something like this:
- Take a screenshot of the reference section
- Go to your favorite LLM, e.g., ChatGPT, copy&paste
the prompt from
prompt-reference-parser.md, attach the screenshot and submit the query. - Copy the output of the LLM to a file
references.json - Run the script and pass the path to
references.jsonas argument
python main.py path/to/references.jsonThe script will go through each reference and compute a score between 0 and 1 based on the closest match it finds on Semantic Scholar. It will then output the information on the command line.