How to transform Terms Documents Matrix into json file in R

Question

I'm trying to get my Terms Documents Matrix which I transformed into a matrix profileM, into a json file. My data looks like:

        |AAZ | AA2 | AAR
--------|----------|---
are     | 0  | 0   |  1
aze     | 1  | 0   |  0
bar     | 0  | 1   |  0
bor     | 1  | 0   |  0
car     | 0  | 1   |  0
dar     | 0  | 0   |  1

Which is:

profileM = matrix( c(0,1,0,1,0,0, 0,0, 1, 0, 1, 0, 1, 0, 0, 0, 0, 1), nrow=6, ncol=3) 
colnames(profileM) <- c("AAZ", "AA2", "AAR")
rownames(profileM) <- c("are", "aze", "bar", "bor", "car","dar")

I want the json file, which retains cells where the value is 1 (or above), to look like this :

{
    "id": AAZ: {

        "text": {
            "aze", "bor"
        }
    },
    "id": AA2: {

        "text": {
            "bar", "car"
         }
    },
     "id": AAR: {

        "text": {
            "are", "dar"
        }
    }

}

I wrote this code:

for(col in 1:ncol(profileM)) {
  x  <- list(id = colnames(profileM)[col])
  x          <- append(x, x$id )
  for(row in 1:nrow(profileM)) {
    if (profileM[row, col] > 0){
      x$text       <- rownames(profileM)[row]
      x        <- append(x, x$text )
    }
  }
}
json <- jsonlite::toJSON(x) 
json

and having this(which is wrong):

{"id":["AAR"],"2":["AAR"],"text":["dar"],"4":["are"],"5":["dar"]}

Could someone help to have a working code, specialy with apply or sapply? Thank you!

Have you seen [this question](https://stackoverflow.com/questions/24840182/converting-a-matrix-dataframe-in-r-to-json-object-with-some-constraints) before? — jazzurro, Feb 06 '18 at 12:57

RLesur · Accepted Answer · 2018-02-06T14:57:53.103

0

Using the tidyverse, you can do:

library(tidyverse)

profileM %>% 
  as_tibble(rownames = "text") %>%
  gather(id, value, -text) %>%
  filter(value == 1) %>% 
  select(-value) %>%
  group_by(id) %>%
  summarise(text = list(text)) %>%
  jsonlite::toJSON()

edited Feb 06 '18 at 14:57

answered Feb 06 '18 at 13:21

RLesur

5,810
1
18
53

@road `value` is a dummy, filtered where it is equal to 1. `id` values are colnames. – RLesur Feb 06 '18 at 17:21
@road did you really execute the code? I don't think – RLesur Feb 06 '18 at 17:32
OK, maybe your dataset isn't a matrix and you need to adapt the first step (the answer uses row names) – RLesur Feb 06 '18 at 17:40

score 0 · Answer 2 · answered Feb 06 '18 at 15:26

You could also avoid having to convert to a matrix and take it straight from a text processing package like quanteda to a dataframe and then convert to json format. Example using inaugural data in quanteda...

library(quanteda)
library(jsonlite)
mycorpus <- corpus_subset(data_corpus_inaugural, Year > 1970)
quantdfm <- dfm(mycorpus, verbose = FALSE, select = "ar*")
quantdfm # you can further refine your criteria right in quanteda
testingdf<-convert(quantdfm, to = "data.frame")
testingdf$id<-row.names(testingdf)
str(testingdf)
toJSON(list(traits = names(testingdf), values = testingdf), pretty = TRUE)

How to transform Terms Documents Matrix into json file in R

2 Answers2