biographR

The biographR package is designed to help extract structured biographical data from unstructured text using ChatGPT’s API.

Installation

You can install the latest version of biographR using the following code:

remotes::install_github("SamuelFrederick/biographR")

Example

Say we have a biography for an individual named John Smith. The biography might look something like:

John Smith graduated from Nowhere College with a B.A. in 1962. 
He then went on to receive a Ph.D. from Nonexistent University. 
At the same time, his wife, Sally Smith, was earning her M.D. from Invisible University.

It would be challenging to extract structured biographical data about John Smith from this text without reading the text and getting the information by hand. Instead, we can use the get_bio() function from the biographR package, which calls ChatGPT’s API to extract the data for us.

library(biographR)
get_bio(bio = "John Smith graduated from Nowhere College with a B.A. in 1962. He then went on to receive a Ph.D. from Nonexistent University. At the same time, his wife, Sally Smith, was earning her M.D. from Invisible University.", 
        bio_name = "John Smith", 
        prompt_fields = c("college", "graduate_school", 
                          "highest_level_of_education", "gender",
                          "married"))
#> Input Tokens: 164
#> Output Tokens: 50
#> Total Tokens: 214
#> # A tibble: 1 × 5
#>   college         graduate_school          highest_level_of_edu…¹ gender married
#>   <chr>           <chr>                    <chr>                  <chr>  <chr>  
#> 1 Nowhere College Nonexistent University;… Ph.D.                  ""     Sally …
#> # ℹ abbreviated name: ¹​highest_level_of_education

We can also customize the output further using the prompt_fields_formats and prompt_fields_values arguments. Additionally, while the package defaults to GPT 3.5, GPT 4 often appears to perform somewhat better in extracting subtler information (e.g., gender from pronouns).

get_bio(bio = "John Smith graduated from Nowhere College with a B.A. in 1962. He then went on to receive a Ph.D. from Nonexistent University. At the same time, his wife, Sally Smith, was earning her M.D. from Invisible University.", 
        bio_name = "John Smith", 
        prompt_fields = c("college", "graduate_school", 
                          "highest_level_of_education", "gender",
                          "married"), 
        prompt_fields_formats = list(college = "{SCHOOL} - {DEGREE}", 
                                    graduate_school = "{SCHOOL} - {DEGREE}", 
                                    highest_level_of_education = "{DEGREE}"), 
        prompt_fields_values = list(married = c("Yes", "No")), 
        openai_model = "gpt-4")
#> Input Tokens: 192
#> Output Tokens: 53
#> Total Tokens: 245
#> # A tibble: 1 × 5
#>   college                graduate_school   highest_level_of_edu…¹ gender married
#>   <chr>                  <chr>             <chr>                  <chr>  <chr>  
#> 1 Nowhere College - B.A. Nonexistent Univ… Ph.D.                  Male   Yes    
#> # ℹ abbreviated name: ¹​highest_level_of_education