A relatively slow version of the Rapid Automatic Keyword Extraction (RAKE) algorithm. See Automatic keyword extraction from individual documents for details on how RAKE works or read the "Getting started" vignette ( vignette("getting-started")).

slowrake(txt, stop_words = smart_words, stop_pos = c("VB", "VBD", "VBG",
  "VBN", "VBP", "VBZ"), word_min_char = 3, stem = TRUE,
  phrase_delims = "[[:space:]]-[[:space:]]|[,.?():;\"!/]|]|\\[")



A character vector, where each element of the vector contains the text for one document.


A vector of stop words which will be removed from your documents. The default value (smart_words) contains the 'SMART' stop words (equivalent to tm::stopwords('SMART')) . Set stop_words = NULL if you don't want to remove stop words.


All words that have a part-of-speech (POS) that appears in stop_pos will be considered a stop word. stop_pos should be a vector of POS tags. All possible POS tags along with their definitions are in the pos_tags data frame (View(slowraker::pos_tags)). The default value is to remove all words that have a verb-based POS (i.e., stop_pos = c("VB", "VBD", "VBG", "VBN", "VBP", "VBZ")). Set stop_pos = NULL if you don't want a word's POS to matter during keyword extraction.


The minimum number of characters that a word must have to remain in the corpus. Words with fewer than word_min_char characters will be removed before the RAKE algorithm is applied. Note that removing words based on word_min_char happens before stemming, so you should consider the full length of the word and not the length of its stem when choosing word_min_char.


Do you want to stem the words before running RAKE?


A regular expression containing the patterns that will be used as phrase delimiters.


An object of class rakelist, which is just a list of data frames (one data frame for each element of txt). Each data frame will have the following columns:


A keyword that was identified by RAKE.


The number of times the keyword appears in the document.


The keyword's score, as per the RAKE algorithm. Keywords with higher scores are considered to be higher quality than those with lower scores.


If you specified stem = TRUE, you will get the stemmed versions of the keywords in this column. When you choose stemming, the keyword's score (score) will be based off its stem, but the reported number of times that the keyword appears (freq) will still be based off of the raw, unstemmed version of the keyword.


slowrake(txt = "some text that has great keywords")
#> #> # A rakelist containing 1 data frames: #> $ :'data.frame': 2 obs. of 4 variables: #> ..$ keyword:"great keywords" ... #> ..$ freq :1 1 #> ..$ score :4 1 #> ..$ stem :"great keyword" ...
slowrake(txt = dog_pubs$title[1:2], stem = FALSE)
#> | | | 0% | |=================================== | 50% | |======================================================================| 100%
#> #> # A rakelist containing 2 data frames: #> $ :'data.frame': 4 obs. of 3 variables: #> ..$ keyword:"assistance dogs" ... #> ..$ freq :1 1 ... #> ..$ score :4 4 ... #> $ :'data.frame': 6 obs. of 3 variables: #> ..$ keyword:"guide dog owners perspectives" ... #> ..$ freq :1 1 ... #> ..$ score :13 ...