Process the abstract with a large language model

Source: R/process_with_ollama.R

A function that batch process data extraction from text.

process_with_ollama(
  input,
  model = "llama3.1:8b",
  type_abstract,
  i,
  abstract_col
)

Arguments

input

A data frame contains abstracts and variables we want to extract. If you are not sure how the data should look like, please see the example dataset abstract. This data should be the testing data saved using the `save_training_data()` function.

model

Large Language Model name you want to use i.e. "llama3.1:8b". If you are not sure what LLMs have been installed locally, please use the function: ellmer:::ollama_models() to find out what LLMs are available.

type_abstract

is created using the type_object function from the ellmer package. Objects represent a collection of named values and are created with type_object(). Objects can contain any number of scalars, arrays, and other objects. They are similar to named lists in R. If still not sure, please look up the documentation of ellmer::type_object().

i

number of abstracts you want to process at once.

abstract_col

column name for the column contains abstract.

Value

Returns a data.frame with results and time.

Examples

library(ellmer)
type_abstract <- type_object("Some key information from abstract.",
  no_patients_llm = type_integer("Find the total number of patients
                                  included in this study.",
                                  required = FALSE),
  no_AKI_llm = type_integer("Number of Acute Kidney Injury (AKI) patients
                             included in this study.",
                             required = FALSE),
  per_AKI_llm = type_number("Percentage of Acute Kidney Injury ",
                             required = FALSE),
  ICU_llm = type_boolean("Included only Intensive Care Unit (ICU) patients.",
                          required = FALSE),
  start_date = type_string("The starting date of the study written in
                            YYYY-MM-DD format",
                            required = FALSE),
  end_date = type_string("The end date of the study written in YYYY-MM-DD
                          format",
                          required = FALSE),
  age_mean_llm = type_number("Find the average age of the study cohort",
                              required = FALSE),
  age_median_llm = type_number("Find the median age of the study cohort",
                                required = FALSE)
)
if (FALSE) { # \dontrun{
process_with_ollama(abstracts,
                    type_abstract = type_abstract,
                    i = 1,
                    abstract_col = "Abstract")
} # }