Module: Lingua::EN::Syllable::Guess

Defined in:
lib/lingua/en/syllable/guess.rb

Overview

Uses english word patterns to guess the number of syllables. A single module method is made available, syllables, which, when passed an english word, will return the number of syllables it estimates are in the word. English orthography (the representation of spoken sounds as written signs) is not regular. The same spoken sound can be represented in multiple different ways in written English (e.g. rough/cuff), and the same written letters can be pronounced in different ways in different words (e.g. rough/bough). As the same series of letters can be pronounced in different ways, it is not possible to write an algorithm which can always guess the number of syllables in an english word correctly. However, it is possible to use frequently recurring patterns in english (such as “a final -e is usually silent”) to guess with a level of accuracy that is acceptable for applications like syllable counting for readability scoring. This module implements such an algorithm. This module is inspired by the Perl Lingua::EN::Syllable module. However, it uses a different (though not larger) set of patterns to compensate for the ‘special cases’ which arise out of English’s irregular orthography. A number of extra patterns (particularly for derived word forms) means that this module is somewhat more accurate than the Perl original. It also omits a number of patterns found in the original which seem to me to apply to such a small number of cases, or to be of dubious value. Testing the guesses against the Carnegie Mellon Pronouncing Dictionary, this module guesses right around 90% of the time, as against about 85% of the time for the Perl module. However, the dictionary contains a large number of foreign loan words and proper names, and so when the algorithm is tested against ‘real world’ english, its accuracy is a good deal better. Testing against a range of samples, it guesses right about 95-97% of the time.

Constant Summary collapse

SubSyl =

special cases - 1 syllable less than expected

[
  /[^aeiou]e$/, # give, love, bone, done, ride ...
  /[aeiou](?:([cfghklmnprsvwz])\1?|ck|sh|[rt]ch)e[ds]$/,
  # (passive) past participles and 3rd person sing present verbs:
  # bared, liked, called, tricked, bashed, matched

  /.e(?:ly|less(?:ly)?|ness?|ful(?:ly)?|ments?)$/,
  # nominal, adjectival and adverbial derivatives from -e$ roots:
  # absolutely, nicely, likeness, basement, hopeless
  # hopeful, tastefully, wasteful

  /ion/, # action, diction, fiction
  /[ct]ia[nl]/, # special(ly), initial, physician, christian
  /[^cx]iou/, # illustrious, NOT spacious, gracious, anxious, noxious
  /sia$/, # amnesia, polynesia
  /.gue$/ # dialogue, intrigue, colleague
]
AddSyl =

special cases - 1 syllable more than expected

[
  /i[aiou]/, # alias, science, phobia
  /[dls]ien/, # salient, gradient, transient
  /[aeiouym]ble$/, # -Vble, plus -mble
  /[aeiou]{3}/, # agreeable
  /^mc/, # mcwhatever
  /ism$/, # sexism, racism
  /(?:([^aeiouy])\1|ck|mp|ng)le$/, # bubble, cattle, cackle, sample, angle
  /dnt$/, # couldn/t
  /[aeiou]y[aeiou]/ # annoying, layer
]
NotUsed =

special cases not actually used - these seem to me to be either very marginal or actually break more stuff than they fix

[
  /^coa[dglx]./, # +1 coagulate, coaxial, coalition, coalesce - marginal
  /[^gq]ua[^auieo]/, # +1 'du-al' - only for some speakers, and breaks
  /riet/, # variety, parietal, notoriety - marginal?
]

Class Method Summary collapse

Class Method Details

.syllables(word) ⇒ Object 



72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
 
# File 'lib/lingua/en/syllable/guess.rb', line 72

def self.syllables(word)
  return 1 if word.length == 1
  word = word.downcase.delete("'")

  syllables = word.scan(/[aeiouy]+/).length

  # special cases
  for pat in SubSyl
    syllables -= 1 if pat.match(word)
  end
  for pat in AddSyl
    syllables += 1 if pat.match(word)
  end

  syllables = 1 if syllables < 1 # no vowels?
  syllables
end