R中带有编码，模式匹配和嘈杂文本的问题 [英] Troubles with encoding, pattern matching and noisy texts in R
问题描述
我们遇到了编码，使用从网上自动下载的文本进行模式匹配的问题.
We are experiencing problems with encoding, pattern matching using texts automatically downloaded from the web.
我们需要一些帮助来了解问题所在以及如何解决.就我个人而言，我必须承认，在阅读了很多关于该主题的文章之后，我完全感到困惑:-)
We need some help to understand where the problem lies and how to fix it. Personally, I must confess that after having read so many posts on the topic, I am completely confused :-)
我们的文字有时包括:1)令人不安的Unicode(我已经读过此书(自动转义unicode字符)，但是我不知道以哪种方式可以帮助正则表达式
Our texts sometimes include: 1) disturbing Unicode (I have read this already (Automatically escape unicode characters ), but I am not sure in which way it can help with regular expressions)
2)奇怪的引号(例如刻度线和双勾号，这是我们无法自动识别的(此页面很有用
2) weird quotes (such as ticks and double ticks than we do not manage to identify automatically (this page is useful https://www.cl.cam.ac.uk/~mgk25/ucs/quotes.html, but how shall we apply this codes within our code?)
我已经应用了这篇文章中给出的建议:如何在RStudio中更改R的语言环境?
I have already applied the suggestions given in this post: How to change the locale of R in RStudio?
为了简而言之，我将提供一个示例(从众多示例中)，这些示例说明了我们正在遇到的问题.
To make a long story short, I will provide an example (out of many) that shows the kind of problems we are experiencing.
阅读下面的代码(简洁但易于阅读):目标是搜索模式"CURRENT URL http://[\ S] *"输入文件"_ansiktstics_corpus.txt"中，并将匹配项保存到输出文件中.
Read the code below (inelegant, but easy to read): the goal is to search the pattern "CURRENT URL http://[\S]*" in the input file "_ansiktstics_corpus.txt" and save the matches onto an output file.
我在底部提供了一个输入文件的片段.
I provide a snippet of an input file at the bottom.
我们的文本中使用的语言是瑞典语.我正在Windows 10上工作.
Language used in our texts is Swedish. I am working on Windows 10.
-----起始代码
library(stringr)
rm(list=ls(all=TRUE)) # clear memory
setwd("yourPath")
seeds_eCare01 <- dir(pattern = "_ansiktstics_corpus.txt") # we have a file list, but for the sake of this example I specify only a file that I attach to allow the reproducibility of the experiment
cat(file="yourPath/urls.csv", append =FALSE)
urlPattern<-"CURRENT URL http://[\\S]*" # all the lines containing the following pattern (we are interested in extracting the URLs)
for (f in seeds_eCare01) {
tempData = scan( f, what="character", encoding = "UTF-8", sep="",quote = NULL)
urlList<-grep("urlPattern", tempData, value = TRUE, perl = TRUE)
# we tried also with "gsub", we get the same error
cat(urlList,file="yourPath/urls.csv", sep="\n",append = TRUE)
}
-----结束代码
控制台输出如下:
-启动控制台输出
Read 13354 items
Warning messages:
1: In grep("urlPattern", tempData, value = TRUE, perl = TRUE) :
input string 18 is invalid UTF-8
2: In grep("urlPattern", tempData, value = TRUE, perl = TRUE) :
input string 19 is invalid UTF-8
3: In grep("urlPattern", tempData, value = TRUE, perl = TRUE) :
input string 4590 is invalid UTF-8
4: In grep("urlPattern", tempData, value = TRUE, perl = TRUE) :
input string 4591 is invalid UTF-8
5: In grep("urlPattern", tempData, value = TRUE, perl = TRUE) :
input string 4593 is invalid UTF-8
--- edn控制台输出
---edn console output
尽管"CURRENT URL"存在，但在文件中未找到任何模式.
No pattern was found in the file, although "CURRENT URL" is there.
我的默认语言环境是:
> Sys.getlocale()
[1] "LC_COLLATE=Swedish_Sweden.1252;LC_CTYPE=Swedish_Sweden.1252;LC_MONETARY=Swedish_Sweden.1252;LC_NUMERIC=C;LC_TIME=Swedish_Sweden.1252"
现在，让我们解决问题:
Now, let’s get to the problem:
我要读取和搜索的文件是使用encoding = UTF-8上传的
The file that I want to read and search is uploaded using encoding=UTF-8
scan( f, what="character", encoding = "UTF-8", sep="",quote = NULL)
但是当我运行以下检查(以检查tempData是否为UTF-8)时，我得到了错误
but when I run the following checks (to check whether the tempData is UTF-8) I get FALSE
all(stri_enc_isutf8(tempData))
[1] FALSE
> stri_enc_mark(tempData)
[1] "ASCII" "ASCII" "ASCII" "ASCII" "ASCII" "ASCII" "ASCII" "ASCII" "ASCII" "ASCII" "UTF-8" "ASCII" "UTF-8" "ASCII" "ASCII" "UTF-8" "UTF-8"
[18] "UTF-8" "UTF-8" "ASCII" "ASCII" "ASCII" "ASCII" "UTF-8" "UTF-8" "ASCII" "ASCII" "ASCII" "ASCII" "ASCII" "ASCII" "ASCII" "ASCII" "ASCII"
我也尝试了"readlines"和连接(如Gries 2017所建议的)，但是我遇到了类似的问题:
I also tried with "readlines" and a connection (as suggested by Gries 2017 )but I experienced similar problems:
(tempData<-readLines(con<-file(f, encoding="UTF-8"), warn=FALSE)); close(con)
当我运行以下命令时，我得到一个TRUE，但grep完全与扫描"失败
when I run the following I get a TRUE, but the grep fails exactly as with "scan"
all(stri_enc_isutf8(tempData))
[1] TRUE
但是，当我运行以下命令(以检查编码)时，我得到了ascii和uft-8的混合物
However, when I run the following (to check the encoding), I get a mixture of ascii and uft-8
stri_enc_mark(tempData)
[1] "ASCII" "ASCII" "ASCII" "ASCII" "ASCII" "ASCII" "UTF-8" "UTF-8" "UTF-8" "ASCII" "ASCII" "UTF-8" "UTF-8" "UTF-8" "UTF-8" "UTF-8" "UTF-8"
问题1)模式匹配不起作用:为什么?2)尽管我们强制使用UTF-8编码，但编码并非始终如一:为什么?3)我们如何摆脱或转换令人不安的字符?任何建议，提示或见解都将不胜感激.
Questions 1) Pattern matching does not work: Why? 2) The encoding is not consistently UTF-8 although we force it: why? 3) How can we get rid or convert disturbing characters ? Any suggestions, hints or insights are greatly appreciated.
预先感谢
Marina干杯
input file: "_ansiktstics_corpus.txt"
<>
<header>
<usergroup>lay/specialized</usergroup>
<annotator's comments>[...]</annotator's comments>
</header>
</>CURRENT URL http://www.aftonbladet.se/halsa/article11873667.ab
"Men det är inte som på film"
Miriam Jaakola, 30, var 26 år när hon fick sin diagnos. Foto: tomas bergman
Hon varken svär eller säger fula ord — men hon drabbas ofta av tics och har svårt att tygla sina
impulser.
Miriam Jaakola, 30, har Tourettes syndrom.
"Jag har haft tics så länge jag kan minnas. När jag var barn hade jag mycket ansiktstics. Jag blinkade, gapade och gjorde grimaser. Jag brukade kalla det för mina ovanor och tänkte egentligen inte så mycket på det. Det var så mycket annat i mitt liv som var rörigt. Jag har en ganska mild form av Tourettes syndrom, så mina tics är inte så tydliga. Det är både på gott och på ont. Folk som inte vet om det märker oftast inte att jag har tourette. Själv förstod jag det inte heller förrän jag var 26 år och fick min diagnos.
Svär inte
Den vanligaste fördomen är att alla som har Tourettes syndrom svär eller säger fula ord. Men ticsen kan se väldigt olika ut. De ändrar sig också över tid. Det är bara en tredjedel av alla med tourette som säger fula ord under någon period i sitt liv. Jag har aldrig gjort det.
Tourette är en sorts känsla – som ett tryck eller en frustration – som byggs upp inom en och till slut bryter ut i tics. Ungefär som när man nyser eller känner att en nysning är på väg. Jag kan hålla tillbaka mina tics om jag är ute bland folk, men då blir de oftast
värre efteråt.Det fi nns en stark energi i tourette. I dag när jag vet hur jag ska hantera det, kan jag vända den energin till något
positivt, till exempel i mitt jobb. Men tourette kan också ställa till problem. Jag har svårt att koncentrera mig och kontrollera impulser. När jag var yngre kunde jag få blixtsnabba utbrott. Jag minns fortfarande första gången jag reflekterade över det. Jag var runt tio år och stod och pillade med något. Plötsligt kastade jag iväg alla grejer. Hade jag haft en allvarligare form av tourette hade jag säkert skrikit eller fått en spasmurladdning. Jag minns att jag tänkte: Undrar om alla har det så här, eller om det bara är jag?
Skoltiden jobbig
Skoltiden var jättejobbig.
推荐答案
Wiktor帮我编写了这段代码.该代码将一个嘈杂的文本语料库转换为一个干净的字符串数据集(.csv)
Wiktor helped me with this code. The code converts a noisy text corpus to a clean dataset of strings(.csv)
rm(list=ls(all=TRUE))
library(NLP)
library(tm)
# Settings
kNonAnnotatedTextsPath <- "path" # The folder path for the text content.
kAnnotatedTextsPath <- "path" # The folder path for the categories.
kOutputPath <- "path" # The destination for the output file.
kOutputFileName <- "output.csv" # The name and format of the output file
kOverwriteOutputFile <- TRUE # Overwrite or append the content to the output file? TRUE = Overwrite, FALSE = Append.
kWarn <- TRUE # Warn the user of uncommon categories.
# Patterns
kStartPattern <- "CURRENT URL" # The text the occur on the line before the text content.
kBreakPattern <- "<>" # The text that occur on the line after the text content.
kCategoryPattern <- ".*<usergroup>(.*)</usergroup>.*" # The text that surrounds the category: (.*)
ExtractCategories <- function(file.name){
# Extracts the categories for a given file. Returns in form of a list vector.
con <- file(paste(kAnnotatedTextsPath, "/", file.name, sep = ""), encoding="UTF-8")
document.sections <- readLines(con, warn=FALSE)
close(con)
document.categories <- vector(mode = "list")
document.names <- c()
for(section in document.sections){
if (grepl(kCategoryPattern, section)){
document.categories <- c(document.categories, gsub(kCategoryPattern, "\\1", section))
}
if (grepl(kStartPattern, section)){
document.names <- c(document.names, section)
}
}
names(document.categories) <- document.names
return(document.categories)
}
ExtractDocuments <- function(file, provided.categories){
# Extracts the text content from a given file, and appends the correct category.
# Returns a list of two, one with a list the text content and one list with the corresponding categories.
collect <- FALSE
con <- file(paste(kNonAnnotatedTextsPath, "/", file, sep = ""), encoding="UTF-8")
document.sections <- readLines(con, warn=FALSE)
close(con)
document.string <- ""
document.list <- c()
document.categories <- c()
document.name <- ""
for(section in document.sections){
if(grepl(kStartPattern, section)){
document.name <- section
collect <- TRUE
} else if(collect == TRUE && grepl(kBreakPattern, section)){
document.categories <- c(document.categories, get(document.name, provided.categories))
document.list <- c(document.list, document.string)
document.name <- ""
document.string <- ""
collect <- FALSE
} else if(collect){
document.string <- paste(document.string, section, sep = " ")
}
}
if(nchar(document.string) != 0){
document.categories <- c(document.categories, get(document.name, provided.categories))
document.list <- c(document.list, document.string)
}
return(cbind(c(document.list), c(document.categories)))
}
RemoveMisc <- function(string){
# Removes the following characters: ", —, –, '
gsub(pattern = "[\\x{201d}\\x{2014}\\x{2013}\\x{02B9}]", replacement = "", string, perl = TRUE)
}
RemoveStartSpace <- function(string){
# Removes space at the start of a paragraph.
gsub(pattern = "^[ ]", replacement = "", string, perl = TRUE)
}
RemoveMultiSpace <- function(string){
# Removes multiple occurances of space in a row,
gsub(pattern = "[ ]{2,}", replacement = " ", string, perl = TRUE)
}
RemoveWebsites <- function(string){
# Removes the common webpage formates from the text.
gsub(pattern = "(?:(?:(?:http[s]*://)|(?:www\\.))+[\\S]*)", replacement = "", string, perl = TRUE)
}
CleanDocuments <- function(documents){
# Cleans the documents of unwanted (combinations of) signs, and replaces uppcarse letters with lowercase.
# Returns the documents as a corpus object.
corpus <- Corpus(VectorSource(documents[, 1]))
meta(corpus, type="indexed", tag="Category") <- documents[, 2]
corpus <- tm_map(corpus, RemoveWebsites)
corpus <- tm_map(corpus, removeNumbers)
corpus <- tm_map(corpus, removePunctuation)
corpus <- tm_map(corpus, RemoveMisc)
corpus <- tm_map(corpus, RemoveStartSpace)
corpus <- tm_map(corpus, RemoveMultiSpace)
corpus <- tm_map(corpus, tolower)
return(corpus)
}
SaveDocuments <- function(corpus, output.file, warn = FALSE){
# Saves the documents to a csv file in the format: '<text>',<category>
counter = 1
while (counter <= length(corpus)){
text <- as.character(corpus[[counter]])
category <- as.character(meta(corpus)[[1]][counter])
if(warn && !(category %in% c("lay", "specialized"))){
print("Warning!")
print(paste("Unusual classification '", category, "'", ", in the following text:", sep = ""))
print(text)
}
padded_text <- paste("'", text, "',", category, sep = "")
write.table(x = padded_text, file = output.file, append = TRUE, sep = "", row.names = FALSE, col.names = FALSE, quote = FALSE)
counter <- counter + 1
}
}
CreateCorpus <- function(overwrite = FALSE){
# Iterates the files and creates the corpus, which is saved as a csv file.
output.file <- paste(kOutputPath, "/", kOutputFileName, sep = "")
seeds <- dir(path = kAnnotatedTextsPath, pattern = "*.txt")
if (overwrite) {
close(file(output.file, open="w"))
}
for (seed in seeds){
document.categories <- ExtractCategories(seed)
document.texts <- ExtractDocuments(seed, document.categories)
corpus <- CleanDocuments(document.texts)
SaveDocuments(corpus, output.file, kWarn)
}
}
CreateCorpus(kOverwriteOutputFile)
这篇关于R中带有编码，模式匹配和嘈杂文本的问题的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！
