Step 2: Build the knowledge database

[VedShh]

Start finetuning model in the sense that make it more specific - make it give better answers, etc.
At this stage we can also start loading it with more papers which will probably increase the efficiency and accuracy.

• Use API to read PDF
• Extract key sections: Abstract, Introduction, Methods, Results, Discussion, and Conclusions (or just remove key references)
• Test the embedding performance of the knowledge database to check if the key information is extracted.