Module: TurkishStemmer

Extended by:
TurkishStemmer
Included in:
TurkishStemmer
Defined in:
lib/turkish_stemmer.rb,
lib/turkish_stemmer/version.rb

Overview

Please note that we use only lowercase letters for all methods. One should normalize input streams before using the ‘stem` method.

Constant Summary collapse

VOWELS = 
"üiıueöao"
CONSONANTS = 
"bcçdfgğhjklmnprsştvyz"
ROUNDED_VOWELS = 
"oöuü"
UNROUNDED_VOWELS = 
"iıea"
FOLLOWING_ROUNDED_VOWELS = 
"aeuü"
FRONT_VOWELS = 
"eiöü"
BACK_VOWELS = 
"ıuao"
AVG_STEMMED_SIZE =

Heuristic size for average Turkish stemmed word size

4
ALPHABET =

Regular expression that checks if the word contains only turkish characters

Regexp.new("^[abcçdefgğhıijklmnoöprsştuüvyz]+$").freeze
NOMINAL_VERB_STATES = 
load_states_or_suffixes("config/nominal_verb_states.yml")
NOMINAL_VERB_SUFFIXES = 
load_states_or_suffixes("config/nominal_verb_suffixes.yml")
NOUN_STATES = 
load_states_or_suffixes("config/noun_states.yml")
NOUN_SUFFIXES = 
load_states_or_suffixes("config/noun_suffixes.yml")
DERIVATIONAL_STATES = 
load_states_or_suffixes("config/derivational_states.yml")
DERIVATIONAL_SUFFIXES = 
load_states_or_suffixes("config/derivational_suffixes.yml")
PROTECTED_WORDS =

Load settings

Protected words

load_settings("protected_words")
LAST_CONSONANT_EXCEPTIONS =

Last consonant exceptions

load_settings("last_consonant_exceptions")
VOWEL_HARMONY_EXCEPTIONS =

Vower harmony exceptions

load_settings("vowel_harmony_exceptions")
SELECTION_LIST_EXCEPTIONS =

Selection list exceptions

load_settings("selection_list_exceptions")
VERSION = 
"0.1.12"

Instance Method Summary collapse

Instance Method Details

#affix_morphological_stripper(word, options = {}) ⇒ Array

A simple algorithm to strip suffixes from a word based on states and transitions.

Parameters:

  • word (String) —

    the word to strip affixes from

  • options (Hash) (defaults to: {}) —

    options for the algorithm

Options Hash (options):

  • :states (Hash) —

    The states and valid transitions

  • :suffixes (Hash) —

    The suffixes with their rules

Returns:

  • (Array) —

    all possible stem versions



377
378
379
380
381
382
383
384
385
386
387
388
389
390
391
392
393
394
395
396
397
398
399
400
401
402
403
404
405
406
407
408
409
410
411
412
413
414
415
416
417
418
419
420
421
422
423
 
# File 'lib/turkish_stemmer.rb', line 377

def affix_morphological_stripper(word, options = {})
  states   = options[:states]   || {}
  suffixes = options[:suffixes] || {}

  return [word] if states.nil?   || states.empty?
  return [word] if suffixes.nil? || suffixes.empty?

  stems    = []
  # Init first state pending transitions
  pendings = generate_pendings(:a, word, states, suffixes)

  while !pendings.empty? do
    transition = pendings.shift
    word       = transition[:word]
    suffix     = suffixes[transition[:suffix]]
    to_state   = states[transition[:to_state]]
    answer     = mark_stem(word, suffix)

    if answer[:stem] == true
      if ENV['DEBUG']
        puts "Word: #{word} \nAnswer: #{answer} \nInfo: #{transition} \nSuffix: #{suffix}"
      end

      if to_state["final_state"] == true
        # We have a valid transition here. It is safe to remove any pendings
        # with the same signature current pending
        remove_pendings_like!(transition, pendings)
        remove_mark_pendings!(pendings)

        stems.push answer[:word]

        unless to_state["transitions"].empty?
          pendings.unshift(*generate_pendings(transition[:to_state], answer[:word], states, suffixes))
        end

      else
        mark_pendings!(transition, pendings)
        pendings.unshift(*generate_pendings(transition[:to_state], answer[:word],
          states, suffixes, mark: true))
      end
    end
  end

  return [word] if pendings.empty? && stems.empty?

  stems.uniq
end

#count_syllables(word) ⇒ Fixnum

Counts syllables of a Turkish word. In Turkish the number of syllables is equals to the number of vowels.

Parameters:

  • word (String) —

    the word to count its syllables

Returns:

  • (Fixnum) —

    the number of syllables



109
110
111
 
# File 'lib/turkish_stemmer.rb', line 109

def count_syllables(word)
  vowels(word).size
end

#derivational_suffix_machine ⇒ Object

Helper method



364
365
366
367
 
# File 'lib/turkish_stemmer.rb', line 364

def derivational_suffix_machine
  affix_morphological_stripper(yield, states: self::DERIVATIONAL_STATES,
    suffixes: self::DERIVATIONAL_SUFFIXES)
end

#generate_pendings(key, word, states, suffixes, options = {}) ⇒ Array

Given a state key and a word, scans through given states and generate valid pending transitions.

Parameters:

  • key (String) —

    the key for states hash

  • word (String) —

    the word to check

  • states (Hash) —

    the states hash

  • suffixes (Hash) —

    the suffixes hash

  • options (Hash) (defaults to: {}) —

    options for pendings

Options Hash (options):

  • :mark (Boolean) —

    Whether this pending is marked for deletion

Returns:

  • (Array) —

    array of pendings

Raises:

  • (ArgumentError) 


247
248
249
250
251
252
253
254
255
256
257
258
259
260
261
262
263
264
 
# File 'lib/turkish_stemmer.rb', line 247

def generate_pendings(key, word, states, suffixes, options = {})
  raise ArgumentError, "State #{key} does not exist" if (state = states[key]).nil?
  mark = options[:mark] || false

  matched_transitions = state["transitions"].select do |transition|
    word.match(/(#{suffixes[transition["suffix"]]["regex"]})$/)
  end

  matched_transitions.map do |transition|
    {
      suffix: transition["suffix"],
      to_state: transition["state"],
      from_state: key,
      word: word,
      mark: mark
    }
  end
end

#has_frontness?(vowel, candidate) ⇒ Boolean

Checks frontness vowel harmony of two vowels according to Turkish vowel harmony.

Parameters:

  • vowel (String) —

    the first vowel

  • candidate (String) —

    the second vowel

Returns:

  • (Boolean)

See Also:



170
171
172
173
174
175
176
177
178
179
180
 
# File 'lib/turkish_stemmer.rb', line 170

def has_frontness?(vowel, candidate)
  return true if vowel.nil? || vowel.empty?
  return true if candidate.nil? || candidate.empty?

  if (FRONT_VOWELS.include?(vowel) && FRONT_VOWELS.include?(candidate)) ||
     (BACK_VOWELS.include?(vowel) && BACK_VOWELS.include?(candidate))
    return true
  end

  false
end

#has_roundness?(vowel, candidate) ⇒ Boolean

Checks roundness vowel harmony of two vowels according to Turkish vowel harmony.

Parameters:

  • vowel (String) —

    the first vowel

  • candidate (String) —

    the second vowel

Returns:

  • (Boolean)

See Also:



151
152
153
154
155
156
157
158
159
160
161
 
# File 'lib/turkish_stemmer.rb', line 151

def has_roundness?(vowel, candidate)
  return true if vowel.nil? || vowel.empty?
  return true if candidate.nil? || candidate.empty?

  if (UNROUNDED_VOWELS.include?(vowel) && UNROUNDED_VOWELS.include?(candidate)) ||
     (ROUNDED_VOWELS.include?(vowel) && FOLLOWING_ROUNDED_VOWELS.include?(candidate))
    return true
  end

  false
end

#has_vowel_harmony?(word) ⇒ Boolean

Checks vowel harmony of a word according to Turkish vowel harmony.

Parameters:

  • word (String) —

    the word to be checked against Turkish vowel harmony

Returns:

  • (Boolean)

See Also:



126
127
128
129
130
131
132
 
# File 'lib/turkish_stemmer.rb', line 126

def has_vowel_harmony?(word)
  word_vowels = vowels(word)
  vowel       = word_vowels[-2]
  candidate   = word_vowels[-1]

  vowel_harmony?(vowel, candidate)
end

#last_consonant!(word) ⇒ String

Transforms a word taken into account last consonant rule.

Parameters:

  • word (String) —

    the word to check for last consonant change

Returns:

  • (String) —

    the changed word



338
339
340
341
342
343
344
345
346
347
348
349
 
# File 'lib/turkish_stemmer.rb', line 338

def last_consonant!(word)
  return word if LAST_CONSONANT_EXCEPTIONS.include?(word)

  consonants  = { 'b' => 'p', 'c' => 'ç', 'd' => 't', 'ğ' => 'k' }
  last_char   = word[-1]

  if consonants.keys.include?(last_char)
    word[-1] = consonants[last_char]
  end

  word
end

#load_settings(key) ⇒ Object

Helper method for loading settings

Parameters:

  • key (String) —

    the key



70
71
72
73
74
75
76
77
78
 
# File 'lib/turkish_stemmer.rb', line 70

def load_settings(key)
  config_path = File.expand_path("../../config/stemmer.yml", __FILE__)

  begin
    YAML.load_file(config_path)[key]
  rescue => e
    raise "Please provide a valid config/stemmer.yml file, #{e}"
  end
end

#load_states_or_suffixes(file) ⇒ Hash

Loads yaml file and symbolizes keys

Parameters:

  • file (String) —

    path to yaml file

Returns:

  • (Hash) —

    the hash with symbols as keys



59
60
61
62
63
64
65
 
# File 'lib/turkish_stemmer.rb', line 59

def load_states_or_suffixes(file)
  config_path = File.expand_path("../../#{file}", __FILE__)

  YAML.load_file(config_path).symbolize_keys
rescue => e
  raise "An error occured loading #{file}, #{e}"
end

#mark_stem(word, suffix) ⇒ Hash

Given a suffix it stems a word according to Turkish orthographic rules

Parameters:

  • word (String) —

    the word to stem

  • suffix (Hash) —

    a suffix record

Returns:

  • (Hash) —

    a stem answer record



271
272
273
274
275
276
277
278
279
280
281
282
283
284
285
286
287
288
289
290
291
292
293
294
295
296
297
298
299
300
301
302
 
# File 'lib/turkish_stemmer.rb', line 271

def mark_stem(word, suffix)
  stem = !PROTECTED_WORDS.include?(word) &&
         (suffix["check_harmony"] &&
         (has_vowel_harmony?(word) || VOWEL_HARMONY_EXCEPTIONS.include?(word))) ||
         !suffix["check_harmony"]

  suffix_applied = suffix["regex"]

  if stem && (match = word.match(/(#{suffix_applied})$/))
    new_word = word.gsub(/(#{match.to_s})$/, '')
    suffix_applied = match.to_s

    if suffix["optional_letter"]
      answer, match = valid_optional_letter?(new_word, suffix["optional_letter"])

      if answer && match
        new_word = new_word.chop
        suffix_applied = match + suffix_applied
      elsif !answer
        new_word = word
        suffix_applied = nil
        stem = false
      end
    end
  else
    stem = false
    suffix_applied = nil
    new_word = word
  end

  { stem: stem, word: new_word, suffix_applied: suffix_applied }
end

#nominal_verbs_suffix_machine ⇒ Object

Helper method. This is just a shortcut.



352
353
354
355
 
# File 'lib/turkish_stemmer.rb', line 352

def nominal_verbs_suffix_machine
  affix_morphological_stripper(yield, states: self::NOMINAL_VERB_STATES,
    suffixes: self::NOMINAL_VERB_SUFFIXES)
end

#noun_suffix_machine ⇒ Object

Helper method. This is just a shortcut.



358
359
360
361
 
# File 'lib/turkish_stemmer.rb', line 358

def noun_suffix_machine
  affix_morphological_stripper(yield, states: self::NOUN_STATES,
    suffixes: self::NOUN_SUFFIXES)
end

#proceed_to_stem?(word) ⇒ Boolean

Checks whether a word can be stemmed or not. This method checks candidate word against nil, protected, length and vowel harmory.

Parameters:

  • word (String) —

    the candidate word for stemming

Returns:

  • (Boolean) —

    whether should proceed to stem or not



187
188
189
190
191
192
193
194
195
196
 
# File 'lib/turkish_stemmer.rb', line 187

def proceed_to_stem?(word)
  if word.nil? || !turkish?(word) ||
    PROTECTED_WORDS.include?(word) ||
    count_syllables(word) <= 1

    return false
  end

  true
end

#stem(original_word) ⇒ String

Stems a Turkish word.

Algorithm consists of 3 parts: pre-process, process and post-process. The pre-process phase is a quick lookup for words that should not be stemmed based on length, protected words list and vowel harmony. The process phase includes a nominal verb suffix and a noun suffix stripper machine. The last phase includes some additional checks and a simple stem selection decision.

Parameters:

  • word (String) —

    the word to stem

Returns:

  • (String) —

    the stemmed word



35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
 
# File 'lib/turkish_stemmer.rb', line 35

def stem(original_word)
  # Preprocess
  return original_word if !proceed_to_stem?(original_word)

  word = original_word.dup

  # Process
  stems = []
  stems << nominal_verbs_suffix_machine { word }
  stems << original_word
  stems.flatten!.uniq!
  stems << stems.map { |word| noun_suffix_machine { word }}
  stems << original_word
  stems.flatten!.uniq!
  stems << stems.map { |word| derivational_suffix_machine { word }}

  # Postprocess
  stem_post_process(stems, original_word)
end

#stem_post_process(stems, original_word) ⇒ String

Post stemming process

Parameters:

  • stems (Array) —

    array of candidate stems

  • original_word (String) —

    the original word

Returns:

  • (String) —

    the stemmed or the original word



203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
 
# File 'lib/turkish_stemmer.rb', line 203

def stem_post_process(stems, original_word)
  if ENV['DEBUG']
    puts "post process for #{original_word}: #{stems}"
  end

  stems = stems.flatten.uniq

  # Reject original word
  stems.reject! { |w| w == original_word }

  # Reject all non-syllable words
  stems.reject! { |w| count_syllables(w) == 0 }

  # Transform last consonant
  stems.map! { |word| last_consonant!(word) }

  # Sort stems by size
  stems.sort! do |x,y|
    if (x.size - AVG_STEMMED_SIZE).abs == (y.size - AVG_STEMMED_SIZE).abs
      x.size <=> y.size
    else
      (x.size - AVG_STEMMED_SIZE).abs <=>  (y.size - AVG_STEMMED_SIZE).abs
    end
  end

  # Check selection list exceptions
  if !(exception = (stems & SELECTION_LIST_EXCEPTIONS)).empty?
    return exception.first
  end

  # Keep first or original word
  stems.empty? ? original_word : stems.first
end

#valid_optional_letter?(word, letter) ⇒ Array

Given a word and a letter it checks if the optional letter can be part of the stem or not.

Examples:

self.valid_optional_letter?("test", "t")
# => [true, 't']

Parameters:

  • word (String) —

    the examined word

  • letter (String) —

    a single letter or a string armed with a regular expression

Returns:

  • (Array) —

    the answer is returned as an array. First element is a Boolean value and second element is the mached character.



315
316
317
318
319
320
321
322
323
324
325
326
327
328
329
330
331
332
 
# File 'lib/turkish_stemmer.rb', line 315

def valid_optional_letter?(word, letter)
  match         = word.match(/(#{letter})$/)
  answer        = true
  matched_char  = nil

  if match
    matched_char  = match.to_s
    previous_char = word[-2]

    answer = if VOWELS.include?(matched_char)
               (previous_char && CONSONANTS.include?(previous_char))
             else
               (previous_char && VOWELS.include?(previous_char))
             end
  end

  [answer, matched_char]
end

#vowel_harmony?(vowel, candidate) ⇒ Boolean

Checks vowel harmony between two vowels

Parameters:

  • vowel (String) —

    the first vowel

  • candidate (String) —

    the second vowel

Returns:

  • (Boolean)

See Also:



140
141
142
 
# File 'lib/turkish_stemmer.rb', line 140

def vowel_harmony?(vowel, candidate)
  has_roundness?(vowel, candidate) && has_frontness?(vowel, candidate)
end

#vowels(word) ⇒ Array

Gets the vowels of a word

Parameters:

  • word (String) —

    the word to get its vowels

Returns:

  • (Array) —

    array of vowels



117
118
119
 
# File 'lib/turkish_stemmer.rb', line 117

def vowels(word)
  word.gsub(/#{CONSONANTS.chars.to_a.join('|')}/,"").chars.to_a
end