- prepare your data
- save papers in
chatpaper/papers
folder - create markdown file for each paper including methods and results text
- create or use existing
prompt_template
intxt
format - update
questions.txt
to use your question list
- save papers in
- update your Xcode to newest version
- run
make init
- copy .env.example to .env and enter your openai key
- run
make chat
, and have fun
- openai-cookbook apps file-q-and-a nextjs
- Get html page if possible
- run script to get method and result sections
- check auto fetched figures and tables
- if the table is not right, try one of these
- open pdf using word, convert word to html
- open pdf using acrobat, select table export as html
- copy table from pdf and paste to excel
- the best option is directly get table from html file
Note: although there exists many kinds of tools to extract content from a PDF document, none of them can do it without introducing errors. Even more, none of the tools can be used to get a consensus result. So if we want to make sure the parsed content doesn't have errors, manually checking the result is necessary.