String extract from multiple columns using a list in R - r

Question

I am trying to extract information from more than 2 columns (2 columns given as an example below) using a list and creating another column which contains the string from the list found from either one of the column specifying which column to look in first. I have the example below and what the desired output is. Hope that helps what I am exactly looking for.
A<-c("This contains NYU", "This has NYU", "This has XT", "This has FIT",
"Something something UNH","I got into UCLA","Hello XT")
B<-c("NYU","UT","USC","FIT","UNA","UCLA", "CA")
data<-data.frame(A,B)
list <- c("NYU","FIT","UCLA","CA","UT","USC")
A B
1 This contains NYU NYU
2 This has NYU UT
3 This has XT USC
4 This has FIT FIT
5 Something something UNH UNA
6 I got into UCLA UCLA
7 Hello XT CA
I would want the code to search from the list and look in column A first and if it cannot find the string then look in column B and if not then give null. By looking at the list, I would like the desired output to look like the below.
A B C
1 This contains NYU NYU NYU
2 This has NYU UT NYU
3 This has XT USC USC
4 This has FIT FIT FIT
5 Something something UNH UNA <NA>
6 I got into UCLA UCLA UCLA
7 Hello XT CA CA

Answer 1

You can transform your list to a regexpr and then apply R regexpr function :
expr <- paste0(list,collapse = "|")
# expr = "NYU|FIT|UCLA|CA|UT|USC" -> Reg expr means NYU or FIT or ......
data[,"C"] <- ""
cols <- rev(names(data)[-(which(names(data)=="C"))])
for(c in cols) {
index <- regexpr(expr,data[,c])
data[,"C"] <- ifelse(index != -1,substr(data[,c],index,index + attr(index,"match.length")-1),data[,"C"])
}
Hope that will helps
Gottavianoni

Answer 2

Another approach could be
#common between column A & vector l
C_tempA <- sapply(df$A, function(x) intersect(strsplit(as.character(x), split = " ")[[1]], l))
#common between column B & vector l
C_tempB <- sapply(df$B, function(x) intersect(as.character(x), l))
#column C calculation
df$C <- ifelse(C_tempA=="character(0)", C_tempB, C_tempA)
df$C[df$C=="character(0)"] <- NA
#final dataframe
df
Output is:
A B C
1 This contains NYU NYU NYU
2 This has NYU UT NYU
3 This has XT USC USC
4 This has FIT FIT FIT
5 Something something UNH UNA NA
6 I got into UCLA UCLA UCLA
7 Hello XT CA CA
Sample data:
df <- structure(list(A = structure(c(4L, 6L, 7L, 5L, 3L, 2L, 1L), .Label = c("Hello XT",
"I got into UCLA", "Something something UNH", "This contains NYU",
"This has FIT", "This has NYU", "This has XT"), class = "factor"),
B = structure(c(3L, 7L, 6L, 2L, 5L, 4L, 1L), .Label = c("CA",
"FIT", "NYU", "UCLA", "UNA", "USC", "UT"), class = "factor")), .Names = c("A",
"B"), row.names = c(NA, -7L), class = "data.frame")
l <- c("NYU","FIT","UCLA","CA","UT","USC")

Answer 3

Use library(tokenizers) from tokenizers package.
Merge two columns and create a new column with merged A and B
data$newC <- paste(data$A, data$B, sep = " " )
Then, follow below loop which will extract values in a vector and then u can cbind the vector in existing dataframe.
newcolumn <- 'X'
for (p in data$newC)
{
if (!is.na(p))
{
x <- which(is.element(unlist(tokenize_words(list, lowercase = TRUE)), unlist(tokenize_words(p, lowercase = TRUE, stopwords = NULL, simplify = FALSE))))
newcolumn <- append(newcolumn,ifelse(x[1]!= 0, list[x[1]], "NA"))
}
}
newcolumn <- newcolumn[-1]
newcolumn
data <- cbind(data, newcolumn)
Hope it helps.
I am getting output of above as what you expected.
Solution Image: