Convert written number to number in R

Does anybody know a function to convert a text representation of a number into an actual number, e.g. 'twenty thousand three hundred and five' into 20305. I have written numbers in dataframe rows and want to convert them to numbers.

In package qdap, you can replace numeric represented numbers with words (e.g., 1001 becomes one thousand one), but not the other way around:

library(qdap)
replace_number("I like 346457 ice cream cones.")
[1] "I like three hundred forty six thousand four hundred fifty seven ice cream cones."

回答1:

Here's a start that should get you to hundreds of thousands.

word2num <- function(word){
    wsplit <- strsplit(tolower(word)," ")[[1]]
    one_digits <- list(zero=0, one=1, two=2, three=3, four=4, five=5,
                       six=6, seven=7, eight=8, nine=9)
    teens <- list(eleven=11, twelve=12, thirteen=13, fourteen=14, fifteen=15,
                  sixteen=16, seventeen=17, eighteen=18, nineteen=19)
    ten_digits <- list(ten=10, twenty=20, thirty=30, forty=40, fifty=50,
                       sixty=60, seventy=70, eighty=80, ninety=90)
    doubles <- c(teens,ten_digits)
    out <- 0
    i <- 1
    while(i <= length(wsplit)){
        j <- 1
        if(i==1 && wsplit[i]=="hundred")
            temp <- 100
        else if(i==1 && wsplit[i]=="thousand")
            temp <- 1000
        else if(wsplit[i] %in% names(one_digits))
            temp <- as.numeric(one_digits[wsplit[i]])
        else if(wsplit[i] %in% names(teens))
            temp <- as.numeric(teens[wsplit[i]])
        else if(wsplit[i] %in% names(ten_digits))
            temp <- (as.numeric(ten_digits[wsplit[i]]))
        if(i < length(wsplit) && wsplit[i+1]=="hundred"){
            if(i>1 && wsplit[i-1] %in% c("hundred","thousand"))
                out <- out + 100*temp
            else
                out <- 100*(out + temp)
            j <- 2
        }
        else if(i < length(wsplit) && wsplit[i+1]=="thousand"){
            if(i>1 && wsplit[i-1] %in% c("hundred","thousand"))
                out <- out + 1000*temp
            else
                out <- 1000*(out + temp)
            j <- 2
        }
        else if(i < length(wsplit) && wsplit[i+1] %in% names(doubles)){
            temp <- temp*100
            out <- out + temp
        }
        else{
            out <- out + temp
        }
        i <- i + j
    }
    return(list(word,out))
}

Results:

> word2num("fifty seven")
[[1]]
[1] "fifty seven"

[[2]]
[1] 57

> word2num("four fifty seven")
[[1]]
[1] "four fifty seven"

[[2]]
[1] 457

> word2num("six thousand four fifty seven")
[[1]]
[1] "six thousand four fifty seven"

[[2]]
[1] 6457

> word2num("forty six thousand four fifty seven")
[[1]]
[1] "forty six thousand four fifty seven"

[[2]]
[1] 46457

> word2num("forty six thousand four hundred fifty seven")
[[1]]
[1] "forty six thousand four hundred fifty seven"

[[2]]
[1] 46457

> word2num("three forty six thousand four hundred fifty seven")
[[1]]
[1] "three forty six thousand four hundred fifty seven"

[[2]]
[1] 346457

I can tell you already that this won't work for word2num("four hundred thousand fifty"), because it doesn't know how to handle consecutive "hundred" and "thousand" terms, but the algorithm can be modified probably. Anyone should feel free to edit this if they have improvements or build on them in their own answer. I just thought this was a fun problem to play with (for a little while).

Edit: Apparently Bill Venables has a package called english that may achieve this even better than the above code.

回答2:

Here's what I think is a better solution.

    library(stringdist)
    library(gdata)
    #Convert numeric words to digits
isNumericWord=function(string, dist=1, method="dl"){
  nums=c("zero","one","two","three","four","five","six","seven","eight","nine",
         "ten","eleven","twelve","thirteen","fourteen","fifteen","sixteen","seventeen","eighteen","nineteen",
         "twenty","thirty","forty","fifty","sixty","seventy","eighty","ninety",
         "hundred","thousand","million","billion","trillion")
  return(any(stringdist(tolower(string),nums,method=method)<=dist))
}
numberTypes=function(string, dist=1, method="dl"){
  nums=c("zero","one","two","three","four","five","six","seven","eight","nine",
         "ten","eleven","twelve","thirteen","fourteen","fifteen","sixteen","seventeen","eighteen","nineteen",
         "twenty","thirty","forty","fifty","sixty","seventy","eighty","ninety",
         "hundred","thousand","million","billion","trillion")
  string=gsub("[[:punct:]]"," ",string)
  wrdsplit=strsplit(string,split=" ")[[1]]
  wrdsplit=wrdsplit[wrdsplit!=""]
  #Handle number types
  wrdsplit=ifelse(stringdist("first",tolower(wrdsplit),method=method)<=dist,"one st",wrdsplit)
  wrdsplit=ifelse(stringdist("second",tolower(wrdsplit),method=method)<=dist,"two nd",wrdsplit)
  wrdsplit=ifelse(stringdist("third",tolower(wrdsplit),method=method)<=dist &
                    tolower(substr(wrdsplit,nchar(wrdsplit),nchar(wrdsplit)))!="y","three rd",wrdsplit)
  wrdsplit=ifelse(stringdist("fourth",tolower(wrdsplit),method=method)<=dist & 
                    tolower(substr(wrdsplit,nchar(wrdsplit),nchar(wrdsplit)))!="y","four th",wrdsplit)
  wrdsplit=ifelse(stringdist("fifth",tolower(wrdsplit),method=method)<=dist & 
                    tolower(substr(wrdsplit,nchar(wrdsplit),nchar(wrdsplit)))!="y","five th",wrdsplit)
  wrdsplit=ifelse(stringdist("sixth",tolower(wrdsplit),method=method)<=dist & 
                    tolower(substr(wrdsplit,nchar(wrdsplit),nchar(wrdsplit)))!="y","six th",wrdsplit)
  wrdsplit=ifelse(stringdist("seventh",tolower(wrdsplit),method=method)<=dist &
                    tolower(substr(wrdsplit,nchar(wrdsplit),nchar(wrdsplit)))!="y","seven th",wrdsplit)
  wrdsplit=ifelse(stringdist("eighth",tolower(wrdsplit),method=method)<=dist &
                    tolower(substr(wrdsplit,nchar(wrdsplit),nchar(wrdsplit)))!="y","eight th",wrdsplit)
  wrdsplit=ifelse(stringdist("ninth",tolower(wrdsplit),method=method)<=dist &
                    tolower(substr(wrdsplit,nchar(wrdsplit),nchar(wrdsplit)))!="y","nine th",wrdsplit)
  wrdsplit=ifelse(stringdist("tenth",tolower(wrdsplit),method=method)<=dist,"ten th",wrdsplit)
  wrdsplit=ifelse(stringdist("twentieth",tolower(wrdsplit),method=method)<=dist,"twenty th",wrdsplit)
  wrdsplit=ifelse(stringdist("thirtieth",tolower(wrdsplit),method=method)<=dist,"thirty th",wrdsplit)
  wrdsplit=ifelse(stringdist("fortieth",tolower(wrdsplit),method=method)<=dist,"forty th",wrdsplit)
  wrdsplit=ifelse(stringdist("fiftieth",tolower(wrdsplit),method=method)<=dist,"fifty th",wrdsplit)
  wrdsplit=ifelse(stringdist("sixtieth",tolower(wrdsplit),method=method)<=dist,"sixty th",wrdsplit)
  wrdsplit=ifelse(stringdist("seventieth",tolower(wrdsplit),method=method)<=dist,"seventy th",wrdsplit)
  wrdsplit=ifelse(stringdist("eightieth",tolower(wrdsplit),method=method)<=dist,"eighty th",wrdsplit)
  wrdsplit=ifelse(stringdist("ninetieth",tolower(wrdsplit),method=method)<=dist,"ninety th",wrdsplit)
  #Handle other number words that end in "th"
  if(length(wrdsplit)>0){
    for(i in 1:length(wrdsplit)){
      substr_end=substr(wrdsplit[i],(nchar(wrdsplit[i])-1),nchar(wrdsplit[i]))
      substr_beg=substr(wrdsplit[i],1,(nchar(wrdsplit[i])-2))
      if(substr_end=="th" & nchar(wrdsplit[i])!=2 & any(stringdist(tolower(substr_beg),nums,method=method)<=dist)){
        wrdsplit[i]=paste(substr_beg, substr_end,sep=" ")
      }
    }
    return(gsub("  "," ",paste(wrdsplit,collapse=" ")))
  }else{
    return("")
  }
}

#Convert number words to digits
Word2Num=function(string, dist=1, method="dl"){
  original=string
  #Define numbers
  one_digits = list(zero=0, one=1, two=2, three=3, four=4, five=5,
                    six=6, seven=7, eight=8, nine=9)
  teens = list(eleven=11, twelve=12, thirteen=13, fourteen=14, fifteen=15,
               sixteen=16, seventeen=17, eighteen=18, nineteen=19)
  ten_digits = list(ten=10, twenty=20, thirty=30, forty=40, fifty=50,
                    sixty=60, seventy=70, eighty=80, ninety=90)
  large_digits = list(hundred=100, thousand=1000, million=1e6, billion=1e9, trillion=1e12)
  double_digits = c(teens,ten_digits)

  #Split the string into words
  string=gsub("-"," ",gsub(" & ", " and ",string,ignore.case=T))
  string=numberTypes(string)
  wrdsplit=strsplit(tolower(string)," ")[[1]]
  wrdsplit=wrdsplit[wrdsplit!=""]
  isNumber=apply(data.frame(wrdsplit),1,isNumericWord)

  #Find groups of numbers
  if(exists("groups")){
    suppressWarnings(rm(groups))
  }
  i=1
  while(i <= length(wrdsplit)){
    if(isNumber[i]==T){
      if(!exists("groups")){
        groups=list(wrdsplit[i])
      }else if(exists("groups")){
        groups=c(groups, wrdsplit[i])
      }
      for(j in (i+1):length(wrdsplit)){
        if(isNumber[j]){
          groups[[length(groups)]]=c(groups[[length(groups)]],wrdsplit[j])
          i=j+1
        }else{
          i=i+1
          break
        }
      }
    }else{
      i=i+1
    }
  }

  #Convert numeric words to numbers
  if(exists("groups")){
    groupNums=groups
    for(j in 1:length(groups)){
      for(i in 1:length(groups[[j]])){
        #If word is a single digit number
        if(any(stringdist(groups[[j]][i],names(one_digits),method=method)<=dist & 
               tolower(substr(groups[[j]][i],nchar(groups[[j]][i]),nchar(groups[[j]][i])))!="y")){
          #If word is a single digit number
          groupNums[[j]][i]=one_digits[stringdist(groups[[j]][i],names(one_digits),method=method)<=dist][[1]]
        }else if(any(stringdist(groups[[j]][i],names(double_digits),method=method)<=dist)){
          #If word is a double digit number
          groupNums[[j]][i]=double_digits[stringdist(groups[[j]][i],names(double_digits),method=method)<=dist][[1]]
        }else if(any(stringdist(groups[[j]][i],names(large_digits),method=method)<=dist)){
          #If word is a large digit number
          groupNums[[j]][i]=large_digits[stringdist(groups[[j]][i],names(large_digits),method=method)<=dist][[1]]
        }
      }
    }

    #Convert the separated numbers to a single number
    defscipen=options("scipen")[[1]]
    options(scipen=999)
    for(i in 1:length(groups)){
      if(length(groupNums[[i]])==1){
        groupNums[[i]]=as.numeric(groupNums[[i]][1])
      }else{
        while(length(groupNums[[i]])>=2){
          if(nchar(groupNums[[i]][2])>nchar(groupNums[[i]][1])){
            #If the next word has more digits than the current word, multiply them
            temp=as.numeric(groupNums[[i]][1])*as.numeric(groupNums[[i]][2])
          }else if(nchar(groupNums[[i]][2])<nchar(groupNums[[i]][1])){
            #if the next word has less digits than the current word, add them
            temp=as.numeric(groupNums[[i]][1])+as.numeric(groupNums[[i]][2])
          }
          #Combine the results
          if(length(groupNums[[i]])>2){
            groupNums[[i]]=c(temp, groupNums[[i]][3:length(groupNums[[i]])])
          }else{
            groupNums[[i]]=temp
          }
        }
      }
    }
    #Recreate the original string
    groupNums=lapply(groupNums, as.character)
    options(scipen=defscipen)
    for(i in 1:length(groups)){
      wrdsplit[which(wrdsplit==groups[[i]][1])]=groupNums[[i]][1]
      if(length(groups[[i]]>1)){
        wrdsplit[which(wrdsplit==groups[[i]][2:length(groups)])]=""
      }
    }
    #Combine numbers with their endings
    wrdsplit=wrdsplit[wrdsplit!=""]
    if(any(wrdsplit[which(wrdsplit %in% unlist(groupNums))+1] %in% c("rd","th","st","nd"))){
      locs=which(wrdsplit %in% unlist(groupNums))
      for(i in length(locs):1){
        wrdsplit[locs[i]]=paste(wrdsplit[c(locs[i],(locs[i]+1))],collapse="")
        wrdsplit=wrdsplit[-(locs[i]+1)]
      }
    }
    return(trim(paste(wrdsplit,collapse=" ")))
  }else{
    return(original)
  }
}