text2vec (version 0.6)

tokenizers: Simple tokenization functions for string splitting

Description

Few simple tokenization functions. For more comprehensive list see tokenizers package: https://cran.r-project.org/package=tokenizers. Also check stringi::stri_split_*.

Usage

word_tokenizer(strings, ...)
char_tokenizer(strings, ...)
space_tokenizer(strings, sep = " ", xptr = FALSE, ...)
postag_lemma_tokenizer(strings, udpipe_model, tagger = "default",
  tokenizer = "tokenizer", pos_keep = character(0),
  pos_remove = c("PUNCT", "DET", "ADP", "SYM", "PART", "SCONJ", "CCONJ",
  "AUX", "X", "INTJ"))

Arguments

strings

character vector

...

other parameters (usually not used - see source code for details).

sep

character, nchar(sep) = 1 - split strings by this character.

xptr

logical tokenize at C++ level - could speed-up by 15-50%.

udpipe_model

- udpipe model, can be loaded with ?udpipe::udpipe_load_model

tagger

"default" - tagger parameter as per ?udpipe::udpipe_annotate docs.

tokenizer

"tokenizer" - tokenizer parameter as per ?udpipe::udpipe_annotate docs.

pos_keep

character(0) specifies which tokens to keep. character(0) means to keep all of them

pos_remove

c("PUNCT", "DET", "ADP", "SYM", "PART", "SCONJ", "CCONJ", "AUX", "X", "INTJ") - which tokens to remove. character(0) is equal to not remove any.

Value

list of character vectors. Each element of list contains vector of tokens.

Examples

# NOT RUN {
doc = c("first  second", "bla, bla, blaa")
# split by words
word_tokenizer(doc)
#faster, but far less general - perform split by a fixed single whitespace symbol.
space_tokenizer(doc, " ")
# }