Instructing R to find variable name in rows when reading csv file

This will get you started but you still have quite a bit of work to do. This works for one line (and it may work for the other two in the example) but it can be extrapolated to work with all of the lines (lapply FTW). Basically you need to rebuild the JSON structure from that single field (there may be alternative ways, especially if you do not need all the fields). It's easier than it might otherwise be since the Yelp schema is known.

You have to attack it in a pretty deterministic way, converting some fields before others, accounting for spaces in field names, dealing with arrays & nested structures, etc. As I said, you have quite a bit of work ahead of you. If your regex-fu is weak, this will provide ample practice to become a regex ninja.

library(stringi)
library(stringr)
library(jsonlite)

txt <- 'business_id: vcNAWiLM4dR7D2nwwJ7nCA, full_address: 4840 E Indian School Rd\nSte 101\nPhoenix, AZ 85018, close: 17:00, open: 08:00, open: true, categories: [Doctors, Health & Medical], city: Phoenix, review_count: 9, name: Eric Goldberg, MD, neighborhoods: [], longitude: -111.98375799999999, state: AZ, stars: 3.5, latitude: 33.499313000000001, attributes: By Appointment Only: true, type: business'
txt <- gsub("\n", "|", txt)

txt <- sub("business_id: ([[:alnum:]\\:]+)", '"business_id": "\\1"', txt)

txt <- sub('attributes: ', '"attributes": {', txt)
txt <- sub('By Appointment Only: ', '"By Appointment Only": ', txt)
txt <- sub('Accepts Credit Cards: ', '"Accepts Credit Cards": ', txt)
txt <- sub('Good For Groups: ', '"Good For Groups": ', txt)
txt <- sub('Outdoor Seating: ', '"Outdoor Seating": ', txt)
txt <- sub('Price Range: ', '"Price Ranges": ', txt)

txt <- sub("full_address: ([[:alnum:][:space:],\\|\\-\\.]+), close:", '"full_address": "\\1", close:', txt)
txt <- sub("full_address: ([[:alnum:][:space:],\\|\\-\\.]+), open:",  '"full_address": "\\1", open:', txt)

txt <- sub("name: (.*), neighborhoods:", '"name": "\\1", "neighborhoods":', txt)

txt <- gsub("open: ([[:alnum:]\\:]+)", '"open": "\\1"', txt)
txt <- sub("close: ([[:alnum:]\\:]+)", '"close": "\\1"', txt)

txt <- sub("longitude: ([[:digit:]\\.-]+)", '"longitude": "\\1"', txt)
txt <- sub("latitude: ([[:digit:]\\.-]+)", '"latitude": "\\1"', txt)

txt <- sub("review_count: ([[:digit:]\\.]+)", '"review_count": "\\1"', txt)
txt <- sub("stars: ([[:digit:]\\.]+)", '"stars": "\\1"', txt)
txt <- sub("state: ([[:alpha:]]+)", '"state": "\\1"', txt)
txt <- sub("city: ([[:alpha:] \\.-]+)", '"city": "\\1"', txt)

txt <- sub("type: ([[:alpha:]]+)", '"type": "\\1"', txt)

cats <- paste0(sprintf('"%s"', str_trim(str_split(str_match_all(txt, "categories: \\[([[:alpha:] &-,]+)\\],")[[1]][,2], ",")[[1]])), collapse=", ")
txt <- sub("categories: \\[([[:alpha:] &-,]+)\\],", '"categories": [' %s+% cats %s+% '],', txt)

txt <- "{" %s+% txt %s+% "}}"

fromJSON(txt)
## $business_id
## [1] "vcNAWiLM4dR7D2nwwJ7nCA"
## 
## $full_address
## [1] "4840 E Indian School Rd|Ste 101|Phoenix, AZ 85018"
## 
## $close
## [1] "17:00"
## 
## $open
## [1] "08:00"
## 
## $open
## [1] "true"
## 
## $categories
## [1] "Doctors"          "Health & Medical"
## 
## $city
## [1] "Phoenix"
## 
## $review_count
## [1] "9"
## 
## $name
## [1] "Eric Goldberg, MD"
## 
## $neighborhoods
## list()
## 
## $longitude
## [1] "-111.98375799999999"
## 
## $state
## [1] "AZ"
## 
## $stars
## [1] "3.5"
## 
## $latitude
## [1] "33.499313000000001"
## 
## $attributes
## $attributes$`By Appointment Only`
## [1] TRUE
## 
## $attributes$type
## [1] "business"

And, whomever gave you this file deserves whatever evil comes their way in their programmatic life. I'd give them back whatever they wanted from this in gnarly XML with EBCDIC encoding.