The folder brca_3.5_turbo contains summaries of all TCGA BRCA text reports

We used Chain of thought prompting i.e Rolling out the answers in multiple calls for better results
  - Ask for the main points first; then ask to summarize
  - Inject cancer subtype into the summary


In the text files, GPT's expanded output is denoted as "expanded version" and final short summary as "final output"

Example: 


prompt1: Our first question to GPT

	Extract information from breast pathology report. List the histological classification, i.e. type of cancer or DCIS,  …. List out and expand the main points. 
		The report is <report_text>

GPT  : The breast pathology report indicates the following information:
	Classification: <>
	Tumor grade: <> ….

prompt2: Please generate a succinct report in 75 words from the above information. Exclude any filler words or sentences. If something is not mentioned or specified, exclude it from the report. Low verbosity

GPT:  A basal subtype infiltrative duct carcinoma mass (2.4 cm x 2.2cm x 2.2cm) with a high mitotic index ....

