Trim tokens using frequency threshold-based feature selection

Returns a tokens object reduced in size based on document and term frequency, usually in terms of a minimum frequency, but may also be in terms of maximum frequencies. Setting a combination of minimum and maximum frequencies will select features based on a range.

Usage

tokens_trim(
  x,
  min_termfreq = NULL,
  max_termfreq = NULL,
  termfreq_type = c("count", "prop", "rank", "quantile"),
  min_docfreq = NULL,
  max_docfreq = NULL,
  docfreq_type = c("count", "prop", "rank", "quantile"),
  max_n = NULL,
  padding = FALSE,
  verbose = quanteda_options("verbose")
)

Arguments

x

a dfm object

min_termfreq, max_termfreq

minimum/maximum values of feature frequencies across all documents, below/above which features will be removed

termfreq_type

how min_termfreq and max_termfreq are interpreted. "count" sums the frequencies; "prop" divides the term frequencies by the total sum; "rank" is matched against the inverted ranking of features in terms of overall frequency, so that 1, 2, ... are the highest and second highest frequency features, and so on; "quantile" sets the cutoffs according to the quantiles (see quantile()) of term frequencies.

min_docfreq, max_docfreq

minimum/maximum values of a feature's document frequency, below/above which features will be removed

docfreq_type

specify how min_docfreq and max_docfreq are interpreted. "count" is the same as [docfreq](x, scheme = "count"); "prop" divides the document frequencies by the total sum; "rank" is matched against the inverted ranking of document frequency, so that 1, 2, ... are the features with the highest and second highest document frequencies, and so on; "quantile" sets the cutoffs according to the quantiles (see quantile()) of document frequencies.

max_n

the maximum number of features. Features are selected based on their frequencies, but the earlier elements in featnames or types are chosen when the frequencies are equal.

padding

if TRUE, leave an empty string where the removed tokens previously existed.

verbose

if TRUE print the number of tokens and documents before and after the function is applied. The number of tokens does not include paddings.

Value

A tokens object with reduced size.

See also

dfm_trim()

Examples

toks <- tokens(data_corpus_inaugural)

# keep only words occurring >= 10 times and in >= 2 documents
tokens_trim(toks, min_termfreq = 10, min_docfreq = 2, padding = TRUE)
#> Tokens consisting of 60 documents and 4 docvars.
#> 1789-Washington :
#>  [1] ""       "of"     "the"    "Senate" "and"    "of"     "the"    ""      
#>  [9] "of"     ""       ":"      ""      
#> [ ... and 1,525 more ]
#> 
#> 1793-Washington :
#>  [1] "Fellow"   "citizens" ","        "I"        "am"       "again"   
#>  [7] "called"   "upon"     "by"       "the"      "voice"    "of"      
#> [ ... and 135 more ]
#> 
#> 1797-Adams :
#>  [1] "When"  "it"    "was"   "first" ""      ","     "in"    "early" "times"
#> [10] ","     "that"  "no"   
#> [ ... and 2,565 more ]
#> 
#> 1801-Jefferson :
#>  [1] ""       "and"    "Fellow" ""       ":"      ""       "upon"   "to"    
#>  [9] ""       "the"    "duties" "of"    
#> [ ... and 1,911 more ]
#> 
#> 1805-Jefferson :
#>  [1] ""             ","            "fellow"       "citizens"     ","           
#>  [6] "to"           "that"         ""             "which"        "the"         
#> [11] "Constitution" "requires"    
#> [ ... and 2,368 more ]
#> 
#> 1809-Madison :
#>  [1] ""          "to"        ""          "from"      "examples"  "of"       
#>  [7] "the"       "most"      ""          "authority" ","         "I"        
#> [ ... and 1,249 more ]
#> 
#> [ reached max_ndoc ... 54 more documents ]

# keep only words occurring >= 10 times and no more than 90% of the documents
tokens_trim(toks, min_termfreq = 10, max_docfreq = 0.9, docfreq_type = "prop",
            padding = TRUE)
#> Tokens consisting of 60 documents and 4 docvars.
#> 1789-Washington :
#>  [1] ""       ""       ""       "Senate" ""       ""       ""       ""      
#>  [9] ""       ""       ":"      ""      
#> [ ... and 1,525 more ]
#> 
#> 1793-Washington :
#>  [1] "Fellow"   "citizens" ""         ""         "am"       "again"   
#>  [7] "called"   "upon"     ""         ""         "voice"    ""        
#> [ ... and 135 more ]
#> 
#> 1797-Adams :
#>  [1] "When"  ""      "was"   "first" ""      ""      ""      "early" "times"
#> [10] ""      ""      ""     
#> [ ... and 2,565 more ]
#> 
#> 1801-Jefferson :
#>  [1] ""       ""       "Fellow" ""       ":"      ""       "upon"   ""      
#>  [9] ""       ""       "duties" ""      
#> [ ... and 1,911 more ]
#> 
#> 1805-Jefferson :
#>  [1] ""             ""             "fellow"       "citizens"     ""            
#>  [6] ""             ""             ""             ""             ""            
#> [11] "Constitution" "requires"    
#> [ ... and 2,368 more ]
#> 
#> 1809-Madison :
#>  [1] ""          ""          ""          ""          "examples"  ""         
#>  [7] ""          "most"      ""          "authority" ""          ""         
#> [ ... and 1,249 more ]
#> 
#> [ reached max_ndoc ... 54 more documents ]