Match Sentences in R - r

Question

I have 2 tables. Table1 is a smaller table with around 10K values.
Table 1 (Sample):
KeyWords PageView
Phillips Trimmer 123
Buy Samsung Mobile 45
Ripe yellow Banana 63
Pepsi 140
Table 2 contains 1 Million Values.
Table 2 (Sample):
KeyWords PageView
Electric Trimmer 123
Samsung Mobile 45
Yellow Ripe Banana 63
Samsung S6 304
Banana 105
Phillips 209
Trimmer Phillips 29
Now I want to take all words from table 1 and look into table 2 an find the best match. The order of words should not have a big influence in the match i.e "Ripe yellow Banana" should match perfectly with "Yellow Ripe Banana". "Buy Samsung Mobile" should match with "Samsung Mobile" and with "Samsung S6".
The final output should look like this.
Table 3:
Word PageView Match
Phillips Trimmer 123 Trimmer Phillips
Buy Samsung Mobile 45 Samsung Mobile
Ripe yellow Banana 63 Yellow Ripe Banana
Pepsi 140 NA
Would really appreciate if we can Stem and Tokenize the sentence before doing a match.
I have tried the following but its not working properly and the loop takes quite a bit of time.
file_1$match <- ""
for(i in 1:dim(file_1)[1]) {
print(i)
x <- grep(file_1$Keywords[i],file_2$Keyword,value = T, ignore.case = m
T,useBytes = T)
x <- paste0(x,"")
file_1$match[i] <- x
}
I have tried using 'agrep' as well varying the 'max.distance' parameter. The results were not as expected.

Answer 1

Edited: I used the "apply" function to perform the following on every row of tab1:
the function inside "apply", takes x[1] which is keyword (let's say "Ripe Yellow Banana"), strsplit splits it by space ("Ripe" "Yellow" "Banana"), sapply performs grepl on each of these splits to see if the pattern exists in tab2. so you will have 3 column of true falses for "Ripe" "Yellow" "Banana". next step is to count the number of trues for each row and output the tab2 with that row number. I also put a if statement to give NA's if the max number of trues is 0:
tab1<-data.frame(Keyword=c("Phillips Trimmer",
"Buy Samsung Mobile","Ripe Yellow Banana","Pepsi"),
PageView=c(123,45,63,140))
tab2<-data.frame(Keyword=c("Electric Trimmer","Samsung Mobile",
"Yellow Ripe Banana","Samsung S6","Banana",
"Phillips","Trimmer Phillips","Buy Trimmer Philips"),
PageView=c(123,45,63,304,105,209,29,21))
tab2$StrLen<-apply(tab2,1,function(x)length(unlist(strsplit(x[1], " "))))
tab1$BestMatch<-apply(tab1,1,function(x){
a <-sapply(unlist(strsplit(x[1], " ")), grepl, tab2$Keyword)
a<-cbind(a,TRUECnt=rowSums(a==TRUE))
a<-as.data.frame(a)
a$StrLen <- length(unlist(strsplit(x[1], " ")))
if (max(a$TRUECnt)==0){
return(NA)
}
return(as.character(tab2[which(a$TRUECnt==max(a$TRUECnt) &
tab2$StrLen <= a$StrLen),]$Keyword))
})
View(tab1) View(tab1)
# Keyword PageView BestMatch
# 1 Phillips Trimmer 123 Trimmer Phillips
# 2 Buy Samsung Mobile 45 Samsung Mobile
# 3 Ripe Yellow Banana 63 Yellow Ripe Banana
# 4 Pepsi 140 <NA>