The article explains how to scrape text from PDF documents and summarize it using OpenAI's Large Language Models (LLMs) in R, highlighting a case study with Meta's 2024 10K Financial Statement. It also promotes a workshop for data scientists on using ChatGPT for data analysis. Entities: OpenAI, Meta, CVS Health, JP Morgan, S&P Global Tags: PDF Scraping, Data Science, OpenAI, Unstructured Data https://www.r-bloggers.com/2024/03/how-to-scrape-pdf-text-and-summarize-it-with-openai-llms-in-r/ -------------- How to Scrape PDF Text and Summarize It with OpenAI LLMs (in R) r-bloggers.com • 2024-03-31T09:00:00-06:00 Companies: OpenAI, Meta, CVS Health, JP Morgan, S&P Global Tags: PDF Scraping, Data Science, OpenAI, Unstructured Data The article discusses how businesses can utilize unstructured data, particularly from PDF documents, by scraping text and summarizing it using OpenAI's Large Language Models (LLMs) in R. It outlines a two-step process for extracting text from PDFs and summarizing it, using Meta's 2024 10K Financial Statement as a case study. Additionally, it promotes a free workshop on using ChatGPT for data scientists. -------------- I want to use this one but with you.com so that I can summarize a pdf as part of the email that get's sent to Dots. Here is the R code so far: # Load Libraries library(fs) library(tidyverse) library(glue) library(blastula) library(RDCOMClient) # Get all files from all subdirectories from the parent directory parent_directory <- "W:/PATACCT/BusinessOfc/Revenue Cycle Analyst/Payer_Policies" all_files <- dir_ls(parent_directory, recurse = TRUE, type = "file") folders <- dir_ls(parent_directory, type = "directory") # Email to send to anthem_address <- "" uhc_address <- "" # Make a tibble of all_files all_files_tbl <- as_tibble(all_files) all_files_tbl <- all_files_tbl |> mutate( file_name = path_file(all_files), ins_name = map(all_files_tbl$value, \(x) x |> strsplit(split = "/") |> pluck(1) |> pluck(6) ) |> unlist(), file_extension = path_ext(all_files), file_size = file_size(all_files), file_date = file_info(all_files)$modification_time, email_body = md(glue( " ## Payer Policy Files Please see attached file {file_name}, from {ins_name} with the file extension {file_extension} and size {file_size} bytes. The file attached is a payer policy. Thank you, The Patient Financial Services Team " )) ) all_files_tbl <- all_files_tbl |> group_by(file_name) |> mutate(rn = row_number()) |> ungroup() |> filter(rn == 1) # Compose Email ---- # Open Outlook # purr the emails out to DoTadda Dots walk( .x = all_files_tbl$email_body, ~ { Outlook <- COMCreate("Outlook.Application") Email <- Outlook$CreateItem(0) Email[["subject"]] <- "Payer Policy Files" Email[["body"]] <- .x attachment <- str_replace_all( all_files_tbl$value[all_files_tbl$email_body == .x], "/", "\\\\" ) Email[["body"]] <- .x Email[["to"]] <- # make a switch statement to send to either address if (str_detect(attachment, "Anthem")) { anthem_address } else if (str_detect(attachment, "UHC")) { uhc_address } else { stop("No email address found") } Email[["attachments"]]$Add(attachment) Email$Send() rm(Outlook) rm(Email) Sys.sleep(10) } )