R break corpus into sentences

Question

1. I have a number of PDF documents, which I have read into a corpus with library tm. How can one break the corpus into sentences?

2. It can be done by reading the file with readLines followed by sentSplit from package qdap [*]. That function requires a dataframe. It would also would require to abandon the corpus and read all files individually.

3. How can I pass function sentSplit {qdap} over a corpus in tm? Or is there a better way?.

Note: there was a function sentDetect in library openNLP, which is now Maxent_Sent_Token_Annotator - the same question applies: how can this be combined with a corpus [tm]?

Answer 1

This is a function built off this Python solution that allows some flexibility in that the lists of prefixes, suffixes, etc. can be modified to your specific text. It's definitely not perfect, but could be useful with the right text.

caps = "([A-Z])"
prefixes = "(Mr|St|Mrs|Ms|Dr|Prof|Capt|Cpt|Lt|Mt)\\."
suffixes = "(Inc|Ltd|Jr|Sr|Co)"
acronyms = "([A-Z][.][A-Z][.](?:[A-Z][.])?)"
starters = "(Mr|Mrs|Ms|Dr|He\\s|She\\s|It\\s|They\\s|Their\\s|Our\\s|We\\s|But\\s|However\\s|That\\s|This\\s|Wherever)"
websites = "\\.(com|edu|gov|io|me|net|org)"
digits = "([0-9])"

split_into_sentences <- function(text){
  text = gsub("\n|\r\n"," ", text)
  text = gsub(prefixes, "\\1<prd>", text)
  text = gsub(websites, "<prd>\\1", text)
  text = gsub('www\\.', "www<prd>", text)
  text = gsub("Ph.D.","Ph<prd>D<prd>", text)
  text = gsub(paste0("\\s", caps, "\\. "), " \\1<prd> ", text)
  text = gsub(paste0(acronyms, " ", starters), "\\1<stop> \\2", text)
  text = gsub(paste0(caps, "\\.", caps, "\\.", caps, "\\."), "\\1<prd>\\2<prd>\\3<prd>", text)
  text = gsub(paste0(caps, "\\.", caps, "\\."), "\\1<prd>\\2<prd>", text)
  text = gsub(paste0(" ", suffixes, "\\. ", starters), " \\1<stop> \\2", text)
  text = gsub(paste0(" ", suffixes, "\\."), " \\1<prd>", text)
  text = gsub(paste0(" ", caps, "\\."), " \\1<prd>",text)
  text = gsub(paste0(digits, "\\.", digits), "\\1<prd>\\2", text)
  text = gsub("...", "<prd><prd><prd>", text, fixed = TRUE)
  text = gsub('\\.”', '”.', text)
  text = gsub('\\."', '\".', text)
  text = gsub('\\!"', '"!', text)
  text = gsub('\\?"', '"?', text)
  text = gsub('\\.', '.<stop>', text)
  text = gsub('\\?', '?<stop>', text)
  text = gsub('\\!', '!<stop>', text)
  text = gsub('<prd>', '.', text)
  sentence = strsplit(text, "<stop>\\s*")
  return(sentence)
}

test_text <- 'Dr. John Johnson, Ph.D. worked for X.Y.Z. Inc. for 4.5 years. He earned $2.5 million when it sold! Now he works at www.website.com.'
sentences <- split_into_sentences(test_text)
names(sentences) <- 'sentence'
df_sentences <- dplyr::bind_rows(sentences) 

df_sentences
# A tibble: 3 x 1
sentence                                                     
<chr>                                                        
1 Dr. John Johnson, Ph.D. worked for X.Y.Z. Inc. for 4.5 years.
2 He earned $2.5 million when it sold!                         
3 Now he works at www.website.com.  

Answer 2

openNLP had some major changes. The bad news is it looks very different than it used to. The good news is that it's more flexible and the functionality you enjoyed before is still there, you just have to find it.

This will give you what you're after:

?Maxent_Sent_Token_Annotator

Just work through the example and you'll see the functionality you're looking for.

Answer 3

I don't know how to reshape a corpus but that would be a fantastic functionality to have.

I guess my approach would be something like this:

Using these packages

# Load Packages
require(tm)
require(NLP)
require(openNLP)

I would set up my text to sentences function as follows:

convert_text_to_sentences <- function(text, lang = "en") {
  # Function to compute sentence annotations using the Apache OpenNLP Maxent sentence detector employing the default model for language 'en'. 
  sentence_token_annotator <- Maxent_Sent_Token_Annotator(language = lang)

  # Convert text to class String from package NLP
  text <- as.String(text)

  # Sentence boundaries in text
  sentence.boundaries <- annotate(text, sentence_token_annotator)

  # Extract sentences
  sentences <- text[sentence.boundaries]

  # return sentences
  return(sentences)
}

And my hack of a reshape corpus function (NB: you will lose the meta attributes here unless you modify this function somehow and copy them over appropriately)

reshape_corpus <- function(current.corpus, FUN, ...) {
  # Extract the text from each document in the corpus and put into a list
  text <- lapply(current.corpus, Content)

  # Basically convert the text
  docs <- lapply(text, FUN, ...)
  docs <- as.vector(unlist(docs))

  # Create a new corpus structure and return it
  new.corpus <- Corpus(VectorSource(docs))
  return(new.corpus)
}

Which works as follows:

## create a corpus
dat <- data.frame(doc1 = "Doctor Who is a British science fiction television programme produced by the BBC. The programme depicts the adventures of a Time Lord—a time travelling, humanoid alien known as the Doctor. He explores the universe in his TARDIS (acronym: Time and Relative Dimension in Space), a sentient time-travelling space ship. Its exterior appears as a blue British police box, a common sight in Britain in 1963, when the series first aired. Along with a succession of companions, the Doctor faces a variety of foes while working to save civilisations, help ordinary people, and right wrongs.",
                  doc2 = "The show has received recognition from critics and the public as one of the finest British television programmes, winning the 2006 British Academy Television Award for Best Drama Series and five consecutive (2005–10) awards at the National Television Awards during Russell T Davies's tenure as Executive Producer.[3][4] In 2011, Matt Smith became the first Doctor to be nominated for a BAFTA Television Award for Best Actor. In 2013, the Peabody Awards honoured Doctor Who with an Institutional Peabody \"for evolving with technology and the times like nothing else in the known television universe.\"[5]",
                  doc3 = "The programme is listed in Guinness World Records as the longest-running science fiction television show in the world[6] and as the \"most successful\" science fiction series of all time—based on its over-all broadcast ratings, DVD and book sales, and iTunes traffic.[7] During its original run, it was recognised for its imaginative stories, creative low-budget special effects, and pioneering use of electronic music (originally produced by the BBC Radiophonic Workshop).",
                  stringsAsFactors = FALSE)

current.corpus <- Corpus(VectorSource(dat))
# A corpus with 3 text documents

## reshape the corpus into sentences (modify this function if you want to keep meta data)
reshape_corpus(current.corpus, convert_text_to_sentences)
# A corpus with 10 text documents

My sessionInfo output

> sessionInfo()
R version 3.0.1 (2013-05-16)
Platform: x86_64-w64-mingw32/x64 (64-bit)

locale:
  [1] LC_COLLATE=English_United Kingdom.1252  LC_CTYPE=English_United Kingdom.1252    LC_MONETARY=English_United Kingdom.1252 LC_NUMERIC=C                           
[5] LC_TIME=English_United Kingdom.1252    

attached base packages:
  [1] stats     graphics  grDevices utils     datasets  methods   base     

other attached packages:
  [1] NLP_0.1-0     openNLP_0.2-1 tm_0.5-9.1   

loaded via a namespace (and not attached):
  [1] openNLPdata_1.5.3-1 parallel_3.0.1      rJava_0.9-4         slam_0.1-29         tools_3.0.1  

Answer 4

I implemented the following code to solve the same problem using the tokenizers package.

# Iterate a list or vector of strings and split into sentences where there are
# periods or question marks
sentences = purrr::map(.x = textList, function(x) {
  return(tokenizers::tokenize_sentences(x))
})

# The code above will return a list of character vectors so unlist
# to give you a character vector of all the sentences
sentences = unlist(sentences)

# Create a corpus from the sentences
corpus = VCorpus(VectorSource(sentences))

Answer 5

The error is meant to be connected with ggplot2 package and the annotate function gives this error, detach the ggplot2 package and then try again. Hopefully it should work.

Answer 6

Just convert your corpus into a dataframe and use regular expressions to detect the sentences.

Here is a function that uses regular expressions to detect sentences in a paragraph and returns each individual sentence.

chunk_into_sentences <- function(text) {
      break_points <- c(1, as.numeric(gregexpr('[[:alnum:] ][.!?]', text)[[1]]) + 1)
      sentences <- NULL
      for(i in 1:length(break_points)) {
        res <- substr(text, break_points[i], break_points[i+1]) 
        if(i>1) { sentences[i] <- sub('. ', '', res) } else { sentences[i] <- res }
      }
      sentences <- sentences[sentences=!is.na(sentences)]
      return(sentences)
    }

...Using one paragraph inside a corpus from the tm package.

text <- paste('Lorem Ipsum is simply dummy text of the printing and typesetting industry. Lorem Ipsum has been the industry standard dummy text ever since the 1500s, when an unknown printer took a galley of type and scrambled it to make a type specimen book. It has survived not only five centuries, but also the leap into electronic typesetting, remaining essentially unchanged. It was popularised in the 1960s with the release of Letraset sheets containing Lorem Ipsum passages, and more recently with desktop publishing software like Aldus PageMaker including versions of Lorem Ipsum.')
mycorpus <- VCorpus(VectorSource(text))
corpus_frame <- data.frame(text=unlist(sapply(mycorpus, `[`, "content")), stringsAsFactors=F)

Use as follows:

chunk_into_sentences(corpus_frame)

Which gives us:

[1] "Lorem Ipsum is simply dummy text of the printing and typesetting industry."                                                                                                                                     
[2] "Lorem Ipsum has been the industry standard dummy text ever since the 1500s, when an unknown printer took a galley of type and scrambled it to make a type specimen book."                                       
[3] "It has survived not only five centuries, but also the leap into electronic typesetting, remaining essentially unchanged."                                                                                       
[4] "It was popularised in the 1960s with the release of Letraset sheets containing Lorem Ipsum passages, and more recently with desktop publishing software like Aldus PageMaker including versions of Lorem Ipsum."

Now with a larger corpus

text1 <- "Lorem Ipsum is simply dummy text of the printing and typesetting industry. Lorem Ipsum has been the industrys standard dummy text ever since the 1500s, when an unknown printer took a galley of type and scrambled it to make a type specimen book. It has survived not only five centuries, but also the leap into electronic typesetting, remaining essentially unchanged. It was popularised in the 1960s with the release of Letraset sheets containing Lorem Ipsum passages, and more recently with desktop publishing software like Aldus PageMaker including versions of Lorem Ipsum."
text2 <- "It is a long established fact that a reader will be distracted by the readable content of a page when looking at its layout. The point of using Lorem Ipsum is that it has a more-or-less normal distribution of letters, as opposed to using 'Content here, content here', making it look like readable English. Many desktop publishing packages and web page editors now use Lorem Ipsum as their default model text, and a search for 'lorem ipsum' will uncover many web sites still in their infancy. Various versions have evolved over the years, sometimes by accident, sometimes on purpose (injected humour and the like)."
text3 <- "There are many variations of passages of Lorem Ipsum available, but the majority have suffered alteration in some form, by injected humour, or randomised words which don't look even slightly believable. If you are going to use a passage of Lorem Ipsum, you need to be sure there isn't anything embarrassing hidden in the middle of text. All the Lorem Ipsum generators on the Internet tend to repeat predefined chunks as necessary, making this the first true generator on the Internet. It uses a dictionary of over 200 Latin words, combined with a handful of model sentence structures, to generate Lorem Ipsum which looks reasonable. The generated Lorem Ipsum is therefore always free from repetition, injected humour, or non-characteristic words etc."
text_list <- list(text1, text2, text3)
my_big_corpus <- VCorpus(VectorSource(text_list))

Use as follows:

lapply(my_big_corpus, chunk_into_sentences)

Which gives us:

$`1`
[1] "Lorem Ipsum is simply dummy text of the printing and typesetting industry."                                                                                                                                     
[2] "Lorem Ipsum has been the industrys standard dummy text ever since the 1500s, when an unknown printer took a galley of type and scrambled it to make a type specimen book."                                      
[3] "It has survived not only five centuries, but also the leap into electronic typesetting, remaining essentially unchanged."                                                                                       
[4] "It was popularised in the 1960s with the release of Letraset sheets containing Lorem Ipsum passages, and more recently with desktop publishing software like Aldus PageMaker including versions of Lorem Ipsum."

$`2`
[1] "It is a long established fact that a reader will be distracted by the readable content of a page when looking at its layout."                                                             
[2] "The point of using Lorem Ipsum is that it has a more-or-less normal distribution of letters, as opposed to using 'Content here, content here', making it look like readable English."     
[3] "Many desktop publishing packages and web page editors now use Lorem Ipsum as their default model text, and a search for 'lorem ipsum' will uncover many web sites still in their infancy."

$`3`
[1] "There are many variations of passages of Lorem Ipsum available, but the majority have suffered alteration in some form, by injected humour, or randomised words which don't look even slightly believable."
[2] "If you are going to use a passage of Lorem Ipsum, you need to be sure there isn't anything embarrassing hidden in the middle of text."                                                                     
[3] "All the Lorem Ipsum generators on the Internet tend to repeat predefined chunks as necessary, making this the first true generator on the Internet."                                                       
[4] "It uses a dictionary of over 200 Latin words, combined with a handful of model sentence structures, to generate Lorem Ipsum which looks reasonable."                                                       
[5] "The generated Lorem Ipsum is therefore always free from repetition, injected humour, or non-characteristic words etc."