如何根据命名载体列表进行分类（〜本体）

Question

Simply put, I have a data frame containing in each row an item type:

df <- data.frame(
  item = 1:5,
  type = c("apple", "orange", "onion", "lettuce", "chicken")
)

I want to categorize each item into a hierarchically higher category, which is defined by the type, according to a list of possible types for each category. I know all the possible types (or can extract them with df$type %>% levels() ).

1) How should I structure the "ontology"/"dictionary" listing all possible values for each category? I thought about a list of named lists, but I am not sure what would be the best way to do that.

ontology = c(
  "fruit" = c("apple", "orange", "banana"),
  "vegetable" = c("onion", "lettuce", "tomato"),
  "meat" = c("chicken", "beef")
)

2) How should I create a variable category in my data frame categorizing each type?

# Basic attempt...
df %>%
  mutate(category = str_match(type %in% ontology))

Expected result:

df
# item    type  category
#    1   apple     fruit
#    2  orange     fruit
#    3   onion vegetable
#    4 lettuce vegetable
#    5 chicken      meat

Answer 1

Here is a base R method with match , unlist and gsub .

# flatten ontology list to named atomic vector where name is category with added digit
flat <- unlist(ontology)
# match position of df$type in flat ontology, pull out name, and remove numeric digit
df$category <- sub("\\d+$", "", names(flat)[match(df$type, flat)])
df
  item    type  category
1    1   apple     fruit
2    2  orange     fruit
3    3   onion vegetable
4    4 lettuce vegetable
5    5 chicken      meat

Answer 2

You could turn ontology into a lookup table:

library(tidyverse)

df <- data.frame(
  item = 1:5,
  type = c("apple", "orange", "onion", "lettuce", "chicken")
)

lookup <- list(    # use list to avoid suffixes on names
    "fruit" = c("apple", "orange", "banana"),
    "vegetable" = c("onion", "lettuce", "tomato"),
    "meat" = c("chicken", "beef")
) %>% 
    imap(~set_names(rep_along(.x, .y), .x)) %>%    # reverse names and objects
    flatten_chr()    # simplify to character vector

lookup
#>       apple      orange      banana       onion     lettuce      tomato 
#>     "fruit"     "fruit"     "fruit" "vegetable" "vegetable" "vegetable" 
#>     chicken        beef 
#>      "meat"      "meat"

which makes categorizing just a matter of subsetting:

df %>% mutate(category = lookup[type])
#>   item    type  category
#> 1    1   apple     fruit
#> 2    2  orange vegetable
#> 3    3   onion vegetable
#> 4    4 lettuce     fruit
#> 5    5 chicken     fruit